Full opinion text
ANDERSON, Circuit Judge: On November 27, 2004, three soldiers in the United States Army serving in Afghanistan — Lieutenant Colonel Michael McMahon, Chief Warrant Officer Travis Grogan, and Specialist Harley Miller— died when the airplane that was transporting them crashed into the side of a mountain. The plane was owned and operated by defendant-appellant Presidential Airways, Inc. (“Presidential”). The soldiers’ survivors (collectively, “McMahon”) filed a wrongful death action in state court against Presidential and appellants Worldwide Services, LLC; STI Aviation, Inc.; and Air Quest, Inc. (also subsidiaries of Presidential’s parent company, all hereinafter referred to collectively as “Presidential”). After removing the case to the federal district court for the Middle District of Florida, Presidential moved to dismiss, arguing that the suit is barred by the principles of Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1950), the political question doctrine, and the combatant activities exception to the Federal Tort Claims Act (FTCA). The district court denied the motion. Presidential now appeals. McMahon’s complaint alleges that Presidential entered into a contract with the Department of Defense (“DOD”) to provide air transportation and other support services in aid of the military mission in Afghanistan. A Statement of Work (“SOW”) governed the relationship between Presidential and the U.S. military. Presidential agreed to furnish “all fixed-wing aircraft, personnel, equipment, tools, material, maintenance, and supervision necessary to perform Short Take-Off and Landing (STOL) passenger, cargo, or passenger and cargo air transportation services” between various locations in Afghanistan, Uzbekistan, and Pakistan. DOD directed what missions would be flown, when they would be flown, and what passengers and cargo would be carried. The SOW provided that Presidential would appoint a project manager to oversee its performance of the contract. Presidential was required to “develop and implement a commercial quality control plan to ensure safe and reliable air transportation in accordance with FAR 135 and 32 CFR 861.” FAR 135 is Part 135 of the civilian Federal Air Regulations, which governs the “commuter or on-demand operations” of commercial operators. 14 C.F.R. § 135.1. 32 C.F.R. § 861 contains the DOD’s own standards for commercial carriers. This regulation required Presidential to “supervise crew selection,” “ensure the risk associated with all flight operations is reduced to the lowest acceptable level,” “ensure that applicants [for flight crew] are carefully screened,” and “ensure[ ] the proper pairing of aircrews on all flights,” among other duties. 32 C.F.R. § 861.4. The SOW also contained more specific restrictions on Presidential’s operations. It dictated how much rest crews had to receive between flights; described specifications for the planes Presidential was to use; and set out minimum and maximum limits on passengers and cargo per mission. Presidential had the ultimate authority to “refuse to fly any mission for safety reasons” (although each such mission had to be rescheduled). McMahon alleges that on November 27, 2004, Presidential was scheduled to transport the three soldiers from Bagram Airfield in northern Afghanistan to Farah, in western Afghanistan. The aircraft chosen for the flight was a CASA 212-CC, twin-engine, turbroprop, fixed-wing aircraft registered with the FAA. McMahon claims that the flight crew requested a southern departure route from Bagram, but then in fact departed in the opposite direction, on a northerly route. McMahon alleges that the crew then turned west to go to Farah, entered a canyon of rapidly rising terrain, and (while attempting to execute a 180-degree turn) crashed into the face of a 16,580-foot mountain. None aboard survived. The soldiers’ survivors filed suit against Presidential in state court, seeking damages under Florida’s wrongful death statute. Presidential removed the case to the federal district court for the Middle District of Florida under the federal officer removal statute. 28 U.S.C. § 1442(a)(1). The district court denied McMahon’s motion to remand the case to state court. McMahon then filed an amended complaint in the district court alleging that Presidential negligently hired and trained the flight crew, negligently assigned the flight crew, negligently planned the route, negligently equipped the aircraft, and otherwise negligently operated the aircraft. Presidential moved to dismiss the case under Rules 12(b)(1) and 12(b)(6). First, Presidential argued that it was entitled to immunity under the principles of Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1950). That case bars soldiers from suing the United States for injuries they incur incident to their military service. Presidential claimed that it was entitled to the government’s Feres immunity under a theory of derivative sovereign immunity. Presidential also argued that the political question doctrine barred the suit. Finally, Presidential argued that the case should be dismissed under Rule 12(b)(6) because McMahon’s claims were preempted by the FTCA’s combatant activities exception, 28 U.S.C. § 2680(j). In response, McMahon sought to convert the Rule 12(b)(6) motion to one for summary judgment by attaching affidavits and exhibits originally introduced in support of her motion for remand. The district court declined to convert the motion and, on Presidential’s motion, struck the extraneous evidence. The district court then denied Presidential’s motion to dismiss. The court rejected Presidential’s claim to Feres immunity because it concluded that Feres was only available in suits against the federal government or its employees. The court concluded that McMahon’s suit did not present a nonjusticiable political question because it did not yet appear that her tort claims against a private contractor would require the court to examine the judgments or strategy of the United States military. The court also refused to recognize a preemption defense based on the combatant activities exception, finding that preemption was not warranted for suits arising out of services contracts. Presidential has brought this interlocutory appeal of the district court’s denial of its motion to dismiss. We organize this opinion as follows: I. Derivative Feres immunity A. Interlocutory jurisdiction B. Derivative Feres immunity 1. Feres doctrine 2. Derivative sovereign immunity 3. Application of Feres rationales to private contractor agents 4. Some form of immunity may be appropriate for private contractor agents a. Incident-to-service test b. Feres as a basis for private contractor immunity where sensitive military judgments may be involved II. Political question doctrine A. Interlocutory jurisdiction B. Political question doctrine 1. Will the case involve a decision that has been constitutionally committed to another branch? 2. Does the suit involve a lack of judicially discoverable and manageable standards? 3. Other Baker factors III. Preemption based on the combatant activities exception IV. Conclusion We address each issue in turn. I. Derivative Feres immunity A. Interlocutory jurisdiction We first consider our jurisdiction over this interlocutory appeal. Under 28 U.S.C. § 1291, we have “jurisdiction of appeals from all final decisions of the district courts ..., except where a direct review may be had in the Supreme Court.” Normally, an order by the district court is not considered “final” and appealable unless it “ends the litigation on the merits and leaves nothing more for the court to do but execute the judgment.” Digital Equip. Corp. v. Desktop Direct, Inc., 511 U.S. 863, 867, 114 S.Ct. 1992, 1995, 128 L.Ed.2d 842 (1994) (quoting Catlin v. United States, 324 U.S. 229, 233, 65 S.Ct. 631, 633, 89 L.Ed. 911 (1945)). The district court’s pretrial denial of derivative Feres immunity does not qualify as a final judgment'under the normal rule. Presidential instead argues that its appeal is proper under the collateral order doctrine. The collateral order doctrine—-a “practical construction” of the final decision rule—permits appeals from “a small category of decisions that, although they do not end the litigation, must nonetheless be considered ‘final.’ ” Swint v. Chambers County Comm’n, 514 U.S. 35, 42, 115 S.Ct. 1203, 1207-08, 131 L.Ed.2d 60 (1995). An interim decision is appealable as a collateral order only if it “(1) conclusively determines the disputed question, (2) resolves an important issue completely separate from the merits of the action, and (3) is effectively unreviewable on appeal from a final judgment.” Sell v. United States, 539 U.S. 166, 176, 123 S.Ct. 2174, 2182, 156 L.Ed.2d 197 (2003) (punctuation and citation omitted). Decisions that satisfy this test are deemed “too important to be denied review and too independent of the cause itself to require that appellate consideration be deferred until the whole case is adjudicated.” Will v. Hallock, 546 U.S. 345, 349, 126 S.Ct. 952, 957, 163 L.Ed.2d 836 (2006) (punctuation and citation omitted). Presidential argued in the district court that it has immunity under the principles of Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1950). In Feres, the Supreme Court held that the government is immune from claims brought by soldiers for their service-related injuries, despite the waiver of sovereign immunity contained in the Federal Tort Claims Act. Id. at 146, 71 S.Ct. at 159. Presidential argues that it is a common law agent of the federal government, entitled to claim the same immunity from soldiers’ service-related tort suits. Presidential’s claim to derivative Feres immunity qualifies as a collateral order. A party is entitled to a collateral order appeal when it has a substantial claim to a true immunity from suit: i.e., an immunity that not only insulates the party from liability, but also prevents the party from being exposed to discovery and/or trial. Courts have recognized a number of immunities from suit, all of which protect important interests that would be irrevocably lost if the holder of the immunity were erroneously required to stand trial. See, e.g., Puerto Rico Aqueduct & Sewer Auth. v. Metcalf & Eddy, Inc., 506 U.S. 139, 144, 113 S.Ct. 684, 687, 121 L.Ed.2d 605 (1993) (Eleventh Amendment immunity); Mitchell v. Forsyth, 472 U.S. 511, 526-27, 105 S.Ct. 2806, 2815-16, 86 L.Ed.2d 411 (1985) (qualified immunity of federal officer); Nixon v. Fitzgerald, 457 U.S. 731, 743, 102 S.Ct. 2690, 2698, 73 L.Ed.2d 349 (1982) (immunity of President from civil damages). Because immunity from suit entails a right to be free from the burdens of litigation, an erroneous denial cannot be redressed through review of the final judgment, and therefore must be reviewed on interlocutory appeal. Presidential has made out a substantial claim to immunity from suit. The government’s Feres immunity from soldiers’ service-related tort claims is justified, in part, by the need to avoid judicial interference with military discipline and sensitive military judgments. Service-related tort claims are often “the types of claims that, if generally permitted, would involve the judiciary in sensitive military affairs at the expense of military discipline and effectiveness.” United States v. Johnson, 481 U.S. 681, 690, 107 S.Ct. 2063, 2069, 95 L.Ed.2d 648 (1987) (punctuation and citation omitted). The Court has suggested that these interests would be threatened not only by the end result of a lawsuit, but also by the process, including the prospect of “compelled depositions and trial testimony by military officers concerning the details of their military commands.” United States v. Stanley, 483 U.S. 669, 682-83, 107 S.Ct. 3054, 3063, 97 L.Ed.2d 550 (1987). As a result, “[e]ven putting aside the risk of erroneous judicial conclusions (which would becloud military decisionmaking), the mere process of arriving at correct conclusions would disrupt the military regime.” Id. at 683, 107 S.Ct. at 3063. As we discuss more fully below, service-related tort suits against private contractors may sometimes threaten interference with sensitive military decisions. And, pursuant to the reasoning in Stanley, the litigation process itself could conceivably cause intolerable interference with those sensitive military judgments. Presidential has therefore stated a substantial claim to a true immunity from suit, such that an erroneous denial would be “effectively unreviewable on appeal from a final judgment.” Sell, 539 U.S. at 176, 123 S.Ct. at 2182 (punctuation and citation omitted). Presidential’s claim thus satisfies the third collateral order factor. Presidential’s claim also satisfies the second factor. Avoiding judicial interference with sensitive military judgments is clearly an “important” interest. Id. It is also “completely separate from the merits of the action,” id., because whether the suit is service-related (the test for Feres immunity in its usual formulation) does not significantly overlap with the merits of the tort suit. Finally, Presidential’s claim satisfies the first factor. Even if the district court reconsidered its decision after the motion to dismiss but prior to trial, its decision for the time being is “conclusive” because it threatens to expose the contractor to arguably harmful discovery in the interim. Id.) see also Mitchell, 472 U.S. at 527, 105 S.Ct. at 2816. Accordingly, because all three factors are satisfied, we have jurisdiction over the district court’s denial of derivative Feres immunity under the collateral order doctrine. B. Derivative Feres immunity Presidential’s primary argument on appeal is that it was a common law agent of the federal government at the time of the accident, and is therefore entitled to the sovereign immunity the government might have under the Feres doctrine. Presidential’s argument relies on the doctrine of derivative sovereign immunity. The existence and scope of derivative Feres immunity are questions of first impression in this Court. 1. Feres doctrine The Feres doctrine is a judicially created exception to the federal government’s waiver of sovereign immunity for common law torts. In general, the United States has waived its sovereign immunity from state law tort suits “under circumstances where the United States, if a private person, would be liable to the claimant in accordance with the law of the place where the act or omission occurred.” 28 U.S.C. § 1346(b)(1). The Federal Tort Claims Act contains a number of explicit exceptions to this waiver of sovereign immunity, such as the combatant activities exception, 28 U.S.C. § 2680(j), and the discretionary functions exception, 28 U.S.C. § 2680(a). The Supreme Court created another such exception in Feres v. United States, 340 U.S. 135, 71 S.Ct. 153, 95 L.Ed. 152 (1950). In Feres, the Supreme Court held that service members are barred from bringing suit against the federal government for injuries that “arise out of or are in the course of activity incident to [military] service.” Id. at 146, 71 S.Ct. at 159. As a result of the Feres doctrine, soldiers may not recover for their service-related injuries in tort suits against the government, even if the tort suit is not barred by an explicit exception to the FTCA. The Feres doctrine has been controversial, even as applied to the government. See United States v. Johnson, 481 U.S. 681, 703, 107 S.Ct. 2063, 2075, 95 L.Ed.2d 648 (1987) (Scalia, J., dissenting, joined by Brennan, Marshall, & Stevens, JJ.) (describing their perception of “the plain error” of Feres). In Kitowski v. United States, 931 F.2d 1526, 1529 (11th Cir.1991), we noted that “there now appears to be some support on the Supreme Court for overruling Feres.” But, as applied to the government, Feres remains the law. The Supreme Court’s latest formulation of the immunity is contained in United States v. Johnson, 481 U.S. 681, 107 S.Ct. 2063, 95 L.Ed.2d 648 (1987). There the Court held that the doctrine rests on three policies. In the first place, “the relationship between the Government and members of its armed forces is distinctively federal in character.” Johnson, 481 U.S. at 689, 107 S.Ct. at 2068 (punctuation and citation omitted). The “distinctively federal” nature of this relationship means, according to the Court, that the government’s liability to soldiers for service-related accidents must be governed by a uniform rule. Id. Second, the Court held that the uniform rule concerning the government’s liability for service-related injuries to soldiers must be one of no liability. The Court arrived at this conclusion by construing the statutory benefits the government provides to injured service members as a congressional cap on the government’s liability for soldiers’ service-related injuries. See Johnson, 481 U.S. at 690, 107 S.Ct. at 2069 (“[T]he statutory veterans’ benefits provide an upper limit of liability for the Government as to service-connected injuries.”) (punctuation and citation omitted). The uniformity policy dictates a uniform rule, and the cap policy justifies a uniform rule of no liability. As a result, the government is not liable to soldiers for torts that cause service-related accidents, apart from the compensation provided by Congress in the form of statutory benefits. As the third and final rationale supporting the Feres bar, the Court has observed that service-related tort claims are often “the types of claims that, if generally permitted, would involve the judiciary in sensitive military affairs at the expense of military discipline and effectiveness.” Johnson, 481 U.S. at 690, 107 S.Ct. at 2069 (punctuation and citation omitted). This rationale was not part of the original justification for Feres, and is found nowhere in the Feres opinion itself. It was added four years later, in United States v. Brown, 348 U.S. 110, 112, 75 S.Ct. 141, 143, 99 L.Ed. 139 (1954), and was acknowledged as one of the Feres policies in Johnson. The third rationale represents a kind of justici-ability constraint. It recognizes that there are certain service-related tort claims that courts should not adjudicate because of the risk that litigation would interfere with military discipline or sensitive military judgments. In sum, for these three reasons, Feres operates to bar all service-related tort claims brought by soldiers against the government. When a soldier incurs injuries incident to service, the United States is deemed not to have waived its sovereign immunity from suit. As a result, the soldier may not recover in a wide variety of tort suits against the government, ranging from suits based on combat activities, to suits based on training activities, to suits based on medical malpractice in a military hospital, to suits based on slips and falls attributable to the government’s negligence on military bases in the United States during peacetime. We must decide, in the instant case, whether this broad-based immunity extends, in whole or in part, to private military contractors. 2. Derivative sovereign immunity Presidential claims that it is entitled to claim the whole of the government’s Feres immunity under the theory of derivative sovereign immunity. The doctrine of derivative sovereign immunity had its origin in Yearsley v. W.A. Ross Construction Co., 309 U.S. 18, 60 S.Ct. 413, 84 L.Ed. 554 (1940). In this somewhat cryptic decision, the Supreme Court appeared to hold that a common law agent may sometimes share in the sovereign immunity of the United States, because “[t]he action of the agent is the act of the government.” Id. at 22, 60 S.Ct. at 415 (punctuation and citation omitted). Since Yearsley, courts have recognized claims of derivative sovereign immunity in a variety of contexts. See, e.g., Butters v. Vance Int’l, Inc., 225 F.3d 462, 466 (4th Cir.2000) (private contractor entitled to derivative sovereign immunity for following commands of foreign sovereign, Saudi Arabia); Myers v. United States, 323 F.2d 580, 583 (9th Cir.1963); Papagianakis v. The Samos, 186 F.2d 257, 261-62 (4th Cir.1950). We have never upheld a claim of derivative sovereign immunity, although the theory has been presented to us on several occasions. We have, however, imposed a limitation on derivative sovereign immunity, if it in fact exists. In Whitaker v. Harvell-Kilgore Corp., 418 F.2d 1010 (5th Cir.1969), we rejected a claim of derivative sovereign immunity on the ground that the entity claiming it was not a common law agent. Id. at 1015 (“Kilgore was clearly an independent contractor and is not entitled to sovereign immunity.”). Because of Whitaker, to make out a claim of derivative sovereign immunity in this circuit, the entity claiming the immunity must at a bare minimum have been a common law agent of the government at the time of the conduct underlying the lawsuit. See Shaw v. Grumman Aerospace Corp., 778 F.2d 736, 740 (11th Cir.1985). Presidential appears to acknowledge that status as a common law agent is necessary to a claim for derivative sovereign immunity. Presidential argues at length that it in fact was a common law agent at the time of the accident underlying this case. But Presidential acknowledges no other limitation on the extent of derivative sovereign immunity. Instead, Presidential argues that status as a common law agent is a sufficient condition for derivative sovereign immunity. According to Presidential, so long as the government has sovereign immunity with respect to the conduct underlying the suit, and so long as the private contractor is a common law agent of the government, the private contractor shares in the government’s sovereign immunity. Presidential thus argues that, because the government likely would have Feres immunity from suit in this case (the soldiers were almost certainly injured “incident to service”), Presidential is derivatively entitled to that same immunity. The upshot of Presidential’s argument is that common law agents of the government would, like the government, have complete immunity from suits brought by soldiers for service-related injuries. We think that the premise of Presidential’s position is quite clearly wrong. Status as a common law agent is not a sufficient condition for derivative sovereign immunity. If it were, then the law of federal officer immunity would be different in certain respects. A federal officer is simply a specialized common law agent, a servant or employee. If Presidential’s theory were true then, because the United States is immune from suit by default (absent its consent), see Hercules, Inc. v. United States, 516 U.S. 417, 422, 116 S.Ct. 981, 985, 134 L.Ed.2d 47 (1996), federal officers would share in the government’s sovereign immunity whenever it has not been waived. But it is simply not the case that federal officers share in the sovereign’s immunity whenever it has not been waived. Federal officers receive immunity in many situations. But unlike the immunity possessed by the government, that immunity must be affirmatively justified, and does not flow automatically from the government’s retained sovereign immunity. See Westfall v. Erwin, 484 U.S. 292, 295, 108 S.Ct. 580, 583, 98 L.Ed.2d 619 (1988) (“[Ajbsolute immunity for federal officials is justified only when ‘the contributions of immunity to effective government in particular contexts outweigh the perhaps recurring harm to individual citizens.’ ”) (quoting Doe v. McMillan, 412 U.S. 306, 320, 93 S.Ct. 2018, 2028, 36 L.Ed.2d 912 (1973)); Anderson v. Creighton, 483 U.S. 635, 638, 107 S.Ct. 3034, 3038, 97 L.Ed.2d 523 (1987) (qualified immunity for federal officers in Bivens actions justified by need to balance benefit of deterring constitutional violations against “risk that fear of personal monetary liability and harassing litigation will unduly inhibit officials in the discharge of their duties”). In general, the Court has emphasized that a federal officer’s immunity is justified only insofar as it is necessary to protect the officer’s performance of a governmental function. See Westfall, 484 U.S. at 296 n. 3, 108 S.Ct. at 583 n. 3 (“In determining the propriety of shielding an official from suit under the circumstances, this Court has long favored a ‘functional’ inquiry — immunity attaches to particular official functions, not to particular offices .... The adoption of this functional approach reflects the Court’s concern ... that federal officials be granted absolute immunity only insofar as the benefits of immunity outweigh the costs.”). Because the government has immunity by default while an officer’s immunity must be affirmatively justified, there are, unsurprisingly, cases where the government is immune but the federal officer is not. An example from current law demonstrates that official immunity and sovereign immunity are not coextensive. Federal officers generally receive only qualified immunity in a Bivens action. See Anderson, 483 U.S. at 638, 107 S.Ct. at 3038 (noting that “[o]ur cases ... generally provid[e] government officials performing discretionary functions with a qualified immunity, shielding them from civil damages liability as long as their actions could reasonably have been thought consistent with the rights they are alleged to have violated”). This means that federal officers are not immune if they violate clearly established constitutional rights and are sued under Bivens. The government itself, on the other hand, remains immune, even if the right allegedly violated by the officer is clearly established. See Bivens v. Six Unknown Named Agents of Fed. Bureau of Narcotics, 403 U.S. 388, 410, 91 S.Ct. 1999, 2012, 29 L.Ed.2d 619 (1971) (Harlan, J., concurring in the judgment) (“However desirable a direct remedy against the Government might be as a substitute for individual official liability, the sovereign still remains immune to suit.”); FDIC v. Meyer, 510 U.S. 471, 485, 114 S.Ct. 996, 1005, 127 L.Ed.2d 308 (1994) (“[W]e implied a cause of action against federal officials in Bivens in part because a direct action against the Government was not available.”) (emphasis in original); Boda v. United States, 698 F.2d 1174, 1176 (11th Cir.1983) (claim against United States for damages for violation of constitutional rights “barred by the doctrine of sovereign immunity”). As a result, the government may be immune from liability for damages for violating a person’s constitutional rights, even though the officer-agent is not. If Presidential’s position were correct, then this result would not be possible, because the federal officer, as a special kind of common law agent, would be derivatively entitled to the same immunity the government possesses. Presidential’s premise of total derivative immunity for common law agents therefore cannot be correct. Moreover, it is clear that if a federal officer cannot claim complete derivative immunity, neither can a mere common law agent. The result urged by Presidential would mean that, in a Bivens suit, a prison guard employed by the government would have only qualified immunity, while a private contractor who works in the prison but is no more than a common law agent would have absolute immunity (assuming the government has not waived it). That result cannot possibly be the law. Instead, it must be that, just as in the area of official immunity, the immunity of a common law agent must be affirmatively justified. Just as with a federal officer, the immunity of a private party, even if a common law agent, must be carefully tailored to protect the governmental functions the private party is exercising. Predictably, courts have been just as vigilant about tailoring the immunity of a private party to its perceived justification as they have been in tailoring the immunity of federal officers. See, e.g., Mangold v. Analytic Servs., Inc., 77 F.3d 1442, 1447 (4th Cir.1996) (emphasizing that private contractors are entitled to immunity only “in the narrow circumstances where the public interest in efficient government outweighs the costs of granting such immunity”); Austin Mun. Securities, Inc. v. NASD, 757 F.2d 676, 687 (5th Cir.1985) (holding that private companies charged with regulatory duties under Exchange Act are only entitled to immunity “to the extent necessary to permit the proper functioning of the regulatory system”). We must exercise the same vigilance in the instant case. It is thus not enough for Presidential to point to its (alleged) status as a common law agent and the government’s (alleged) Feres immunity. For Presidential to show that it is entitled to the government’s Feres immunity, it must also show that the policies underlying the Feres bar extend to private contractor agents. This conclusion is not contrary to any derivative sovereign immunity case. None of these cases say that status as a common law agent is a sufficient condition for the agent to have the government’s immunity, only that it is a necessary condition. See, e.g., Whitaker, 418 F.2d at 1015. Nor could status as a common law agent be a sufficient condition; otherwise, as discussed above, aspects of the law of official immunity would make little sense. We therefore turn to the policies underlying the Feres doctrine, and examine whether they justify immunity for private contractor agents from all service-related suits brought by soldiers. 3. Application of Feres rationales to private contractor agents To determine whether Presidential, a private contractor, is derivatively entitled to the government’s Feres immunity, we must decide whether and how the policies underlying Feres apply to private contractor agents. The first Feres policy is ensuring that the government faces a uniform rule with respect to injuries incurred by soldiers incident to service. See Johnson, 481 U.S. at 689, 107 S.Ct. at 2068. The Court adopted in Feres a uniform rule of no liability in service-related tort suits brought by soldiers (apart from the statutory benefits provided by the government). See id. While the uniformity rationale does continue to apply to the government, it does not apply at all to private contractors. To apply the rationale to private military contractors would be highly anomalous. Government agencies in general (apart from the military) do not have the benefit of a Feres bar, and therefore must face the non-uniform tort law of the various states (so long as an explicit FTCA exception does not apply). See 28 U.S.C. § 1346(b). For example, the federal prison system must contend with non-uniform tort law. See United States v. Muniz, 374 U.S. 150, 162, 83 S.Ct. 1850, 1857, 10 L.Ed.2d 805 (1963). In Muniz, the Supreme Court expressly refused to extend the Feres uniformity rationale to the federal prison system. See id. (“Without more definite indication of the risks of harm from diversity, we conclude that the prison system will not be disrupted by the application of Connecticut law in one case and Indiana law in another .... ”). Private contractor agents have no more pressing need for a uniform rule. If anything, they have far less exposure than a gigantic federal agency that would face non-uniform rules on a daily basis, and therefore much less risk of having to contend with varying law. The first Feres rationale therefore does not apply to a private contractor such as Presidential. The second policy underlying the Feres bar is the cap on the government’s liability for service-related injuries, set at the amount of statutory benefits provided to the service member. Johnson, 481 U.S. at 690, 107 S.Ct. at 2068. The government’s liability is capped because it has compensated the soldier to some degree (by way of the statutory remedy). The Court has therefore implied that these benefits are all that Congress intended for the government to provide the soldier. See id. The cap policy also does not apply to the private contractor. The private contractor, unlike the government, has not had to pay anything to the soldier, and will not have to pay anything apart from what the soldier might recover in a tort suit. Nor is there any warrant to read the cap on the government’s liability as an implied cap on the private contractor’s. There is absolutely no indication that Congress, in providing statutory benefits for soldiers, intended them to substitute in any way for a remedy against private contractor agents of the military, or intended them to cap the liability of private contractors to soldiers they injure in the course of duty. We therefore decline to recognize the cap policy as a justification for applying the Feres bar to private contractors. It is not surprising that the first two Feres policies apply only to the government, because they serve to protect distinctively sovereign interests — ensuring that the government is not crippled by a non-uniform standard for soldiers’ injuries incurred incident to service, and ensuring that the government’s liability is capped at the amount of statutory benefits it provides to injured soldiers. The Supreme Court has itself implicitly recognized that these two Feres policies do not apply to individuals, such as private contractors. The Supreme Court has twice relied on the Feres policies, in another context, to refuse to create a Bivens cause of action against federal employees for service-related constitutional torts. On both occasions, it did not rely on the first two Feres rationales. See Chappell v. Wallace, 462 U.S. 296, 299, 103 S.Ct. 2362, 2365, 76 L.Ed.2d 586 (1983) (in a case refusing to imply Bivens action against superior officer where the soldier was injured incident to service, relied only on discipline rationale of Feres, not cap or uniformity); United States v. Stanley, 483 U.S. 669, 678-83, 107 S.Ct. 3054, 3061-63, 97 L.Ed.2d 550 (1987) (also relying, post-Johnson, on only the third rationale of Feres in refusing to create Bivens action against government employee where serviceman’s injury was incident to service). Rather, the Court relied solely on the third Feres rationale — preventing interference with military discipline and military judgments. The Court eschewed reliance on the first two Feres rationales even though the defendant in each ease was a government employee — a special kind of common law agent. We therefore do not hesitate to find that these two policies do not apply to private contractors, even if they are agents of the military. On the other hand, we do believe that the third Feres policy potentially has some application to private contractor agents of the military. The third justification for the Feres bar is that it protects against interference with military discipline and sensitive military judgments. Johnson, 481 U.S. at 690, 107 S.Ct. at 2069. We have previously recognized that this rationale embodies two “distinct” concerns: “(1) whether the suit requires the court to second-guess military decisions, and (2) whether the suit might impair essential military discipline.” Shaw, 778 F.2d at 742 (punctuation and citation omitted). We described the military judgments strand as “the classic separation of powers theory.” Id. The “essential military discipline” strand, on the other hand, prevents disruption of the relationship between the service member and his military superiors and cohorts. Id. The discipline strand of the third Feres rationale does not have application in the context of private contractors. In Shaw, we noted that the “essential military discipline” rationale itself embodies two concerns: “(1) the notion that a soldier might use the civilian courts to challenge the act or order of a superior officer; and (2) the idea that in a civilian suit of any sort involving a serviceman, members of the military might be compelled to testify against one another.” Id. at 742. We concluded that neither concern would justify protection of the military contractor. In the first place, the concern that a soldier might use a suit to challenge a superi- or officer is absent because a private contractor is not in the chain of command. Id. at 742-43. That observation is equally true where the private contractor happens to be an agent of the government. Because the private contractor agent is not in the chain of command, a soldier cannot use a suit against the contractor to “challenge the act or order of a superior officer.” Id. at 742. In the second place, we concluded in Shaw that any risk to discipline from the process of trying a case against a private contractor was too remote to be accorded significant weight. Although we acknowledged that soldiers might have to testify on opposite sides in a suit against a private contractor, we stated that “the likelihood of any profound disruption of discipline is negligible from testimony in suits against military contractors.” Shaw, 778 F.2d at 743. We recognize that here too, in a suit against a private contractor agent, there may be soldiers testifying on opposite sides. But we agree with the Shaw Court that this does not present a significant threat to military discipline. Our conclusion that the discipline policy does not have significant application to private contractor agents is fortified by the opinion of the four dissenters in Johnson, who set out several reasons why the discipline policy does not even apply strongly to the government itself. Justice Scalia said, It is strange that Congress’ “obvious” intention to preclude Feres suits because of their effect on military discipline was discerned neither by the Feres Court nor by the Congress that enacted the FTCA (which felt it necessary expressly to exclude recovery for combat injuries). Perhaps Congress recognized that the likely effect of Feres suits upon military discipline is not as clear as we have assumed, but in fact has long been disputed .... Or perhaps Congress assumed that the FTCA’s explicit exclusions would bar those suits most threatening to military discipline, such as claims based upon combat command decisions, 28 U.S.C. § 2680(j); claims based upon performance of “discretionary” functions, § 2680(a); claims arising in foreign countries, § 2680(k); intentional torts, § 2680(h); and claims based upon the execution of a statute or regulation, § 2680(a).... Or perhaps— most fascinating of all to contemplate— Congress thought that barring recovery by servicemen might adversely affect military discipline. After all, the morale of Lieutenant Commander Johnson’s comrades-in-arms will not likely be boosted by news that his widow and children will receive only a fraction of the amount they might have recovered had he been piloting a commercial helicopter at the time of his death. Johnson, 481 U.S. at 699-700, 107 S.Ct. at 2073-74 (Scalia, J., dissenting, joined by three others) (emphasis added). We of course may not deny that the discipline rationale continues to apply to the government, even if the reasoning in the Johnson dissent has some force; the Johnson majority held that it still does. But we are perfectly free to hold that, given the arguably tenuous application of the discipline rationale to the government itself, there is no warrant to extend it to private contractors, a context in which its application is clearly even more tenuous. We are confident in our judgment that there is no substantial impact on military discipline from soldiers recovering for the torts of private contractors, and are free to draw that conclusion because neither this Court nor the Supreme Court has held that it does extend to private contractors. Indeed, in Shaw, we held the opposite. While we do not accept that the discipline portion of the third Feres rationale applies to private contractors, we do recognize that the other part — the risk of a tort suit interfering with sensitive military judgments — does potentially apply to private contractor agents. As the Court in Johnson recognized, suits involving accidents that occurred incident to service may “implicate[] the military judgments and decisions that are inextricably intertwined with the conduct of the military mission.” Johnson, 481 U.S. at 691, 107 S.Ct. at 2069. Applying a tort law standard of care to sensitive military judgments is problematic for two reasons, both of which potentially apply to private contractor agents entrusted with executing or making such judgments. In the first place, there is a problem of institutional competence. Where sensitive military judgments are involved, courts lack the capacity to determine the proper tradeoff between military effectiveness and the risk of harm to the soldiers. For example, it is not possible for a court to develop a standard of care for a reasonably safe infantry assault on a fortified enemy outpost. See Chappell, 462 U.S. at 302, 103 S.Ct. at 2366 (“The complex, subtle, and professional decisions as to the composition, training, equipping, and control of a military force are essentially professional military judgments ....”) (quoting Gilligan v. Morgan, 413 U.S. 1, 10, 93 S.Ct. 2440, 2446, 37 L.Ed.2d 407 (1973)); Aktepe v. United States, 105 F.3d 1400, 1404 (11th Cir.1997) (“[Cjourts lack standards with which to assess whether reasonable care was taken to achieve military objectives while minimizing injury and loss of life.”); Tiffany v. United States, 931 F.2d 271, 279 (4th Cir.1991) (noting that courts cannot develop standards for a prudent intercept). Rather than develop such standards where they lack the expertise, courts instead trust that the military is in the best position to determine the appropriate tradeoff between safety and combat effectiveness. When a private contractor agent is entrusted with making or executing such sensitive military judgments, courts would be similarly powerless to determine whether the agent appropriately balanced military effectiveness and the safety of the soldiers. See Boyle, 487 U.S. at 511, 108 S.Ct. at 2518 (offering as one reason for the military contractor defense the fact that determining liability would involve “the balancing of many technical, military, and even social considerations, including specifically the trade-off between greater safety and greater combat effectiveness”). The court would therefore be equally incapable of entertaining the tort suit. Second, even if courts could determine what a “reasonable bombing” or a “reasonable intercept” would be, it would violate the separation of powers for courts to second-guess the military’s decision in a tort suit. These sorts of sensitive military judgments have been constitutionally committed to the political branches. See Stanley, 483 U.S. at 682, 107 S.Ct. at 3063 (emphasizing “the insistence (evident from the number of Clauses devoted to the subject) with which the Constitution confers authority over the Army, Navy, and militia upon the political branches”); Aktepe, 105 F.3d at 1404 (noting that it “would express a lack of respect for the political branches of government by subjecting their discretionary military and foreign policy decisions to judicial scrutiny, notwithstanding the judiciary’s relative lack of expertise in these areas”). Even if courts were competent to develop liability standards in the area of sensitive military judgments, it would breach separation of powers to apply those standards to the military. It would similarly violate separation of powers for the courts to interfere with sensitive military judgments made or executed by private contractor agents of the military. The military has the constitutionally exclusive authority to make those kinds of judgments, and judicial oversight of the private contractor agents the military uses to execute those judgments would likewise violate separation of powers principles. This sensitive-military-judgments strand of the third Feres rationale embodies concerns about justiciability and separation of powers. It is thus related to the political question doctrine, which is a constitutional restraint on the jurisdiction of the federal courts. See Baker v. Carr, 369 U.S. 186, 217, 82 S.Ct. 691, 710, 7 L.Ed.2d 663 (1962) (stating that factors justifying abstention under political question doctrine include “a lack of judicially discoverable and manageable standards” and “a textually demonstrable constitutional commitment of the issue to a coordinate political department”). Demonstrating this close connection, when the Court has discussed the military judgments portion of the third Feres rationale, it has often mixed in reference to political question cases. See United States v. Shearer, 473 U.S. 52, 58, 105 S.Ct. 3039, 3043, 87 L.Ed.2d 38 (1985) (citing Gilligan v. Morgan, 413 U.S. 1, 93 S.Ct. 2440, 37 L.Ed.2d 407 (1973)); Chappell, 462 U.S. at 301, 103 S.Ct. at 2366 (iciting Gilligan). Our political question cases have, in turn, cited the third Feres rationale when discussing this concern about interfering with military judgments. See Aktepe, 105 F.3d at 1403 (citing Chappell and Stanley). It is evident that the military judgments strand in Feres and the military judgments strand under the political question doctrine overlap and reinforce each other. 4. Some form of immunity may be appropriate for private contractor agents We thus acknowledge that private contractor agents may be entitled to some form of immunity that protects their making or executing sensitive military judgments, and that overlaps and possibly extends beyond the protection provided by the political question doctrine. The question then becomes whether the Feres “incident to service” test is the proper way to protect private contractor agents performing such functions. a. Incident-to-service test We think it quite clear that the incident-to-service test sweeps far too broadly to protect this concern. Some suits barred by the incident-to-service test simply do not involve sensitive military judgments that courts lack the competence or authority to deal with. For example, where a private contractor agent is running a mess hall on a peacetime base, and a soldier gets food poisoning attributable to the contractor’s negligence, the suit would be barred under the “incident to service” test. But in such a suit, there is no concern about interfering with sensitive military judgments. Courts do not lack well-developed standards for dealing with “negligent food service.” Nor does it offend separation of powers for a court to say that a private contractor agent must pay for its negligence in serving tainted food to soldiers. Because “incident to service” necessarily includes a large number of these non-sensitive military judgments, it is far too broad to protect contractors from suits that might interfere with sensitive military judgments. Proof that some suits “incident to service” do not offend separation of powers or present justiciability problems is easily found in the fact that civilians are not barred from bringing some such suits against the government, even though a soldier would be. It is well-established that Feres applies only to soldiers, and therefore does not bar civilian suits along some portion (probably a large portion) of the incident-to-service spectrum. See, e.g., Boyle, 487 U.S. at 510-11, 108 S.Ct. at 2517-18 (Feres “covers only service-related injuries, and not injuries caused by the military to civilians,” and as a result, “it could not be invoked to prevent, for exam-pie, a civilian’s suit against the manufacturer of fighter planes, based on a state tort theory, claiming harm from what is alleged to be needlessly high levels of noise produced by the jet engines”); Taber, 67 F.3d at 1047 (“Feres does not bar suits against the government when the injured plaintiff is a civilian. This remains the case even though the injurer is in the military and military discipline is directly involved.”) (emphasis in original). Civilians are barred from bringing suit against the government only to the extent the government has not waived its sovereign immunity — for example, where the combatant activities exception applies. As a result, a large number of suits on the “incident to service” spectrum can be brought by civilians. This means that some number of “incident to service” suits must not inherently offend separation of powers or present problems of justiciability. If they did, civilians too would be barred from suing the government. But, as Judge Calabresi has observed, “[I]n any number of civilian cases, the alleged judicial inquiry into (and interference with) military affairs, occurs anyway. And if this interference occurs regularly in any event, it cannot possibly raise the constitutional concerns that warrant tortured statutory construction and judicial abstention.” Taber, 67 F.3d at 1047. The fact that civilians are not barred from suing the government along some part of the incident-to-service spectrum indicates that those suits are not inherently non-justiciable, and do not involve sensitive military judgments that are outside judicial competence. The government, of course, does receive immunity in service-related suits brought by soldiers, even when they do not implicate sensitive military judgments. For example, the soldier may not sue the government in the mess hall hypothetical discussed above, if it is the government that is operating the mess hall. But that is because the government’s Feres immunity is supported by policies other than protecting sensitive military judgments: namely, the uniformity, cap, and discipline policies. First, the government needs a uniform rule, even where sensitive military judgments are not implicated. Second, the soldier’s statutory benefit represents the cap on the government’s liability. And third, the suit may impair discipline because it would pit the soldier against his employer and may involve conflict with his superior officer or another soldier. As discussed above, these rationales do not apply to a private contractor. Immunity for private contractors is justified only by the need to protect the making and execution of sensitive military judgments. And, as demonstrated above, a number of “incident to service” suits — probably a substantial number — do not implicate sensitive military judgments, because they can be brought by civilians. As a result, the derivative Feres immunity of private contractors cannot possibly extend to the outer limits of “incident to service.” We therefore hold that private contractors, even if they are agents, are not entitled to immunity from all service-related suits brought by soldiers. b. Feres as a basis for private contractor immunity where sensitive military judgments may be involved As demonstrated above, private contractors are, at most, entitled to a more constricted form of derivative Feres immunity. The Feres “incident to service” test is too broad for the purpose of protecting private contractor agents performing sensitive military functions. But a narrower formulation might conceivably be appropriate for the purpose of protecting this kind of military judgment. It might be said that the political question doctrine, with its explicit concerns regarding justiciability and separation of powers, exhausts the category of military judgments insulated from judicial review. We are, however, willing to entertain the possibility that the political question doctrine, while it informs this strand of Feres, does not necessarily exhaust the category of sensitive military functions that should be protected from judicial scrutiny. We might, for example, accord derivative Feres immunity to a private contractor agent across some spectrum narrower than “incident to service” but possibly broader than political question. This might conceivably include suits involving quintessential or peculiarly military judgments that courts should not hear as a matter of prudence, rather than a matter of constitutional law. Even if such an immunity is warranted, however, we do not believe that the Feres doctrine is an appropriate ground upon which to build it. The reason is that an immunity built on Feres would only prevent soldiers — and would not prevent civilians — from bringing suit against private military contractors making or executing sensitive military judgments. The immunity would necessarily, operate in this way because derivative immunity can be no broader than the sovereign immunity that grounds it, and the government’s Feres immunity only extends to suits brought by soldiers. See, e.g., Boyle, 487 U.S. at 510, 108 S.Ct. at 2517-18 {Feres “covers only service-related injuries, and not injuries caused by the military to civilians”). Feres immunity bars only soldiers — not civilians — even where the civilian suit would require the court to examine a potentially sensitive military judgment. Compare Shearer, 473 U.S. at 58, 105 S.Ct. at 3043 (Feres barred suit by soldier against government for negligent supervision of tortfeasor) with Sheridan v. United States, 487 U.S. 392, 402-03, 108 S.Ct. 2449, 2456, 101 L.Ed.2d 352 (1988) (Feres did not bar suit by civilian against government for negligent supervision of tortfea-sor); see also Taber, 67 F.3d at 1047 (“Feres does not bar suits against the government when the injured plaintiff is a civilian. This remains the case even though the injurer is in the military and military discipline is directly involved.”) (emphasis in original). Because derivative sovereign immunity cannot' be broader than the sovereign immunity that grounds it, this means that a hypothetical sensitive military judgments immunity grounded on Feres could only prevent soldiers from bringing suit. This consequence means that derivative Feres immunity.is an inappropriate vehicle for a “sensitive military judgments” immunity for private contractor agents. In the first place, it would not be effective to protect sensitive military judgments from judicial review because it would not bar suits brought by civilians that implicate such judgments. As an example, any such immunity, as applied to the plane crash in the instant case, would not apply to civilians on board the plane. The employees of Presidential itself (three of whom perished in the accident) could sue Presidential for Presidential’s negligence. Those suits would not be barred by the hypothetical derivative Feres immunity tailored to protect sensitive military judgments. Yet they would present the very same threat of subjecting sensitive military judgments to second-guessing by a court. On top of the ineffectiveness of an immunity grounded on Feres, it would also be inequitable because it would single out soldiers for special disfavor in the courts of law. As noted above, Feres applies only to soldiers, so derivative Feres immunity, even if restricted to sensitive military judgments, could apply only to soldiers as well. But that would have consequences that belie common sense. For example, assume the hypothetical situation of a sensitive military function being performed by a private contractor agent that does not fall within the category of cases barred by the political question doctrine. Assume also three people injured by the contractor’s performance of the sensitive military function: a soldier, a civilian employee of the private contractor, and a journalist. If we extended Feres derivatively to the private contractor, the soldier could not sue the contractor. The employee of the private contractor could sue because, by hypothesis, the suit would not be barred by the political question doctrine. And so could the journalist, for the same reason. There is simply no principled reason why this result should obtain. We refuse to single out soldiers who sacrifice their lives and limbs for our country for special disfavor, even for the laudatory purpose of protecting sensitive military judgments from judicial interference. As a result, we must hold that derivative Feres immunity does not exist in this case. Three of the four recognized Feres rationales do not apply to private contractors. And while protecting sensitive military judgments could conceivably ground an immunity, Feres is an inappropriate vehicle because it would single out soldiers and would not protect sensitive military judgments in suits brought by anyone else (including journalists or private contractor employees). This is not the first time that Feres has been found to be an inappropriate ground upon which to build a protection for private contractors who are involved with sensitive military judgments. In Boyle v. United Technologies Corp., 487 U.S. 500, 108 S.Ct. 2510, 101 L.Ed.2d 442 (1988), the Court held that military contractors have an affirmative defense in a products liability suit where “(1) the United States approved reasonably precise specifications; (2) the equipment conformed to those specifications; and (3) the supplier warned the United States about the dangers in the use of the equipment that were known to the supplier but not to the United States.” Boyle, 487 U.S. at 512, 108 S.Ct. at 2518. But the Court refused to ground this so-called military contractor defense on Feres, for the following reasons: [I]t seems to us that the Feres doctrine, in its application to the present problem, logically produces results that are in some respects too broad and in some respects too narrow. Too broad, because if the Government contractor defense is to prohibit suit against the manufacturer whenever Feres would prevent suit against the Government, then even injuries caused to military personnel by a helicopter purchased from stock ... would be covered. Since Feres prohibits all service-related tort claims against the Government, a contractor defense that rests upon it should prohibit all service-related tort claims against the manufacturer — making inexplicable the three limiting criteria for contractor immunity .... On the other hand, reliance on Feres produces (or logically should produce) results that are in another respect too narrow. Since that doctrine covers only service-related injuries, and not injuries caused by the military to civilians, it could not be invoked to prevent, for example, a civilian’s suit against the manufacturer of fighter planes, based on a state tort theory claiming harm from what is alleged to be needlessly high levels of noise produced by the jet engines. Yet we think that the character of the jet engines the Government orders for its fighter planes cannot be regulated by state tort law, no more in suits by civilians than in suits by members of the Armed Services. Boyle, 487 U.S. at 510-11, 108 S.Ct. at 2517-18. The reasons why the Supreme Court rejected Feres as a basis for the military contractor defense in Boyle are essentially the same reasons why we refuse to ground the hypothesized “sensitive military judgment” immunity on Feres today. The Court in Boyle recognized that reliance on Feres would produce results that are too broad, because the “incident to service” test would include suits against contractors that did not implicate the policies the Court was concerned about in Boyle. Similarly, in this case, an immunity built on Feres would be too broad, because “incident to service” would cover tort suits that do not implicate sensitive military judgments. Then in Boyle, the Court held that a defense grounded on Feres would be too narrow because, even where the policies the Court was concerned about were present, civilians could still bring suit, thus vitiating the purpose of the immunity. Here too, an immunity built on Feres would be too narrow because it would only protect against suits implicating sensitive military judgments that are brought by soldiers, and not against those brought by civilians, or even employees of the private contractor itself. We emphasize that we do not mean to foreclose the possibility of an immunity that is broader than political question but narrower than “incident to service.” We simply hold that any such immunity must, if it is to have any basis in r