OPINION OF THE COURT
SLO VITER, Circuit Judge.I.
The underlying fact situation in this case is similar to that which we considered in Marino v. Bowers, No. 80-1395, 657 F.2d 1363 (3d Cir. Sept. 8, 1981), decided today. An employee of a Pennsylvania county who was hired when his political party was in power was discharged when that party lost power. Thereafter, in Elrod v. Burns, 427 U.S. 347, 96 S.Ct. 2673, 49 L.Ed.2d 547 (1976), the Supreme Court decided patronage dismissals were unconstitutional, and the employee brought suit. Under ordinary circumstances, our holding in the Marino case that Elrod should not be applied retroactively would be determinative of this case as well. Because this case followed a significantly different procedural course than did Marino, it requires separate consideration.
II.
Plaintiff-appellee, John Weaver, was hired in March 1972 as a park superintendent by the Bucks County Board of Commissioners,1 then controlled by a Democratic *1358majority. At all times relevant to this litigation, Weaver was a registered member of the Democratic Party and was active in party politics. In January 1976 Weaver was discharged by the newly elected Republican-controlled board. On November 15, 1977, Weaver filed his complaint against the commissioners, individually and in their official capacities, alleging that his discharge was “solely by reason of his political party affiliation” and violated his civil rights, specifically his First Amendment right to free speech and his Fourteenth Amendment right to due process. The suit was brought under 42 U.S.C. §§ 1983, 1985, 1986, 1988.
Unlike their response in the later-filed Marino case, defendants did not file a motion to dismiss raising the nonretroactivity of the Elrod doctrine. Instead, defendants filed an answer to the complaint denying that Weaver was dismissed because of his political party affiliation and alleging instead that he was dismissed because of his lack of ability and poor work performance. Defendants also denied that Weaver’s position of park superintendent was non-policy-making, thereby seeking to avoid application of the holding in Elrod v. Burns.2
In this posture the ease proceeded to a bench trial before the district court for three days from April 23, 1979 to April 25, 1979. Following the trial, the court issued an opinion on May 17, 1979. After reviewing the evidence, the court found that “all of the evidence in this case points to the inescapable conclusion that plaintiff’s employment was terminated because of his political affiliation only.” The court found “no evidence ... of poor job performance or subordination [sic] on the part of [Weaver]”, and rejected defendants’ argument that Weaver was a policymaking employee. The court stated that Weaver was “nothing more than a glorified foreman, notwithstanding his job title of park superintendent.” The court, relying on the Supreme Court’s holding in Elrod, found the defendants liable. In an order entered the same day, May 17, 1979, the district court directed the defendants to hire Weaver immediately “at the same or similar position and salary which he had at the time of termination, with all of the rights and privileges of seniority which he would have had if the termination had not taken place.” The court declined to enter a judgment for lost wages because Weaver had introduced no evidence to substantiate his claim for lost wages.
There followed a series of post-trial motions and maneuvers which need not be reviewed in detail. Weaver sought modification of the judgment to include the award of backpay which the district court had originally declined to enter, and also moved for attorney’s fees (motions filed May 29, 1979). Before the district court had taken any action on these motions, defendants appealed to this court (June 6, 1979). At a conference with the district court, the district judge apparently indicated he would retain jurisdiction notwithstanding the appeal, and directed the parties to place all outstanding issues before the court. Defendants filed post-trial exceptions to the factual findings of the court (August 13, 1979). This court entered an order remanding the case to the district court to rule on the outstanding post-trial motions (October 12, 1979). Defendants then refiled their post-trial exceptions directed to the district court’s factual findings (October 19, 1979). Weaver filed a motion to hold defendants in contempt because they had not yet rehired him (October 31, 1979).
*1359On November 1, 1979, the district court filed an order denying “all motions which have been filed by the defendants since the date of trial” and scheduling a hearing on Weaver’s contempt motion. On November 27, 1979, the district court ruled on Weaver’s motion to amend the judgment to include a backpay award. The court stated that backpay had not originally been granted because additional information had to be submitted, that Weaver had thereafter submitted the necessary information, and that he was therefore entitled to a backpay award. The court entered judgment in the amount of $40,280.59 for backpay. In a separate opinion the same day, the court calculated that Weaver’s counsel should be awarded attorney’s fees in the amount of $13,200.00 and $1,213.87 for reimbursement of expenses and ordered payment.
On December 6, 1979, defendants filed post-trial exceptions in the form of a motion to amend and modify the judgment in which they contended for the first time (1) that the decision in Elrod v. Burns should be given prospective effect only, and (2) that the decisions of the district court “violate the doctrine of governmental immunity afforded to the defendant by the Constitution of the United States.” Weaver sought to amend the judgment to include backpay and counsel fees for the supplemental period from July 23,1979 to November 26,1979 when he was rehired. By opinion and order filed January 17, 1980, the district court granted Weaver’s request and supplemented the judgment by awarding additional backpay and counsel fees. The court denied defendants’ motion, holding that defendants had waived their claims regarding the nonretroactivity of Elrod and immunity by raising them well after the trial. The court also indicated that the post-trial motion itself was not timely filed.
Defendants appealed, essentially presenting four issues for review: Whether the district court erred (1) in finding that plaintiff was illegally discharged under the standards set forth in Elrod; (2) in applying Elrod retroactively; (3) in refusing to find that defendants were immune; and (4) in dismissing defendants’ immunity and nonretroactivity defenses because not timely raised.
III.
We first address defendants’ contention that the district court erred in finding that Weaver’s discharge fell within the proscription of Elrod v. Burns. We have examined the record and cannot call clearly erroneous the court’s finding that Weaver was dismissed because of his political affiliation. Similarly, even were the relevant inquiry whether Weaver was or was not a policymaking employee, see note 2 supra, the district court’s finding that he was not is also supported by ample evidence, and is not clearly erroneous. Thus, had defendants timely raised their claim that Elrod should not be applied retroactively to cover Weaver’s dismissal, the same considerations which impelled us to decide against retroactivity in Marino v. Bowers, No. 80-1395, 657 F.2d 1363 (3d Cir. Sept. 8, 1981), would be dispositive. Weaver claims, however, that we are precluded from accepting that defense because defendants’ contention was not timely raised.
Defendants’ claim that Elrod should not be applied retroactively attacks the legal sufficiency of the complaint. It is, as such, a defense that the complaint fails to state a claim upon which relief can be granted and should have been made no later than “at the trial on the merits”. Fed. R.Civ.P. 12(h)(2). Under the scheme envisaged by Rule 12(h), challenges to the maintenance of an action are divided into three categories. The least favored category, which encompasses the defenses such as lack of jurisdiction over the person, improper venue, and improper or insufficient service of process, must be raised either before or in a responsive pleading. Fed.R.Civ.P. 12(h)(1). This permits the opposing party to address the defect at the initial stage of *1360the litigation, when it can often be corrected in a timely fashion. In direct contrast, the third category, lack of subject matter jurisdiction, may be raised at any time, Fed.R.Civ.P. 12(h)(3), including on appeal. Trent Realty Associates v. First Federal Savings & Loan Association of Philadelphia, 657 F.2d 29 (3d Cir. 1981). Because this defect deprives the court of its power to adjudicate the controversy, it cannot be waived.
The intermediate category covering the defense at issue here is governed by Rule 12(h)(2) which provides:
A defense of failure to state a claim upon which relief can be granted, a defense of failure to join a party indispensable under Rule 19, and an objection of failure to state a legal defense to a claim may be made in any pleading permitted or ordered under Rule 7(a), or by motion for judgment on the pleadings, or at the trial on the merits.
Although Rule 12(h)(2), in contrast to Rule 12(h)(1), does not expressly provide that the covered defenses will be waived if they are not presented by the time provided, the clear purport of the rule mandates that construction, and it has been so construed. See, e. g., Brule v. Southworth, 611 F.2d 406, 409 (1st Cir. 1979); Simpson v. Providence Washington Insurance Group, 608 F.2d 1171, 1174 (9th Cir. 1979); Snead v. Department of Social Services of New York, 409 F.Supp. 995, 1000 (S.D.N.Y.1975) (three-judge court). See also 5 Wright & Miller, Federal Practice and Procedure § 1392, at 862. A contrary interpretation would effectively negate the clear design of Rule 12(h) to provide a well-defined course which litigants must follow to preserve their defenses.
We do not consider that a claim raised in a post-trial motion, even were that motion itself timely, is made “at the trial on the merits”. While authority is sparse as to what constitutes presenting a defense “at” the trial, it would appear that the defense must be presented so that the court may consider whether there has been a failure to state a claim before disposition on the merits. It is desirable to raise the Rule 12(h)(2) defenses as early as possible, preferably before trial commences if the ruling could avoid any unnecessary trial time. The Rule does not mandate that course, since it permits presentation of any of these defenses “at” the trial. However, if they are raised before the conclusion of the trial, the decisionmaker, whether judge or jury, may be spared the time and effort of reaching a conclusion. As the court stated in Snead v. Department of Social Services of New York, 409 F.Supp. 995, 1000 (S.D.N.Y.1975) (three-judge court):
The clear thrust of the Rule is that a failure to state a claim may be raised at any time before a disposition on the merits but not after. Any other construction would not only add to intolerable delay, but create uncertainty as to the validity of a final judgment on the merits, (emphasis in original).
In Black, Sivalls & Bryson, Inc. v. Shondell, 174 F.2d 587, 590-91 (8th Cir. 1949), the court stated that a defense of failure to state a claim made in defendant’s motion for judgment notwithstanding the verdict was waived because it was made “after the trial and not on the trial.” Similarly, in Jordan v. Weaver, 472 F.2d 985, 995 (7th Cir. 1973), rev’d on other grounds, Edelman v. Jordan, 415 U.S. 651, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974), the court considered precisely the issue before us, i. e., the retro-activity of a Supreme Court constitutional ruling, and stated that because this ground was not presented to the district judge before the entry of judgment, it came too late. It must be noted, however, that the court then proceeded to consider and reject the merits of the defendant’s assertion that the ruling should be accorded only prospective effect. Id. at 995-96.
Therefore, we agree with Weaver that defendants’ claim that Elrod v. Burns should be applied only prospectively was presented untimely because it was not *1361raised until substantially after the conclusion of the trial. We must next determine whether this untimeliness is so fundamental a defect so as to deprive us of the power to consider defendants’ claim. We have found little analysis on this issue elsewhere. We note, however, that there have been cases in which the defense of failure to join an indispensable party has been considered despite the failure of the movant to raise it “at the trial”. See Martinez Moll v. Levitt & Sons of Puerto Rico Inc., 583 F.2d 565 (1st Cir. 1978); Provident Tradesmens Bank & Trust Co. v. Lumbermens Mutual Casualty Co., 365 F.2d 802 (3d Cir. 1966), rev’d on other grounds sub nom. Provident Tradesmens Bank & Trust Co. v. Patterson, 390 U.S. 102, 88 S.Ct. 733, 19 L.Ed.2d 936 (1968); McShan v. Sherrill, 283 F.2d 462 (9th Cir. 1960). While these cases may have analogized the absence of an indispensable party to a lack of subject matter jurisdiction, see 5 Wright & Miller, Federal Practice and Procedure § 1392, at 862, an issue we need not resolve here, Rule 12(h)(2) places the defenses of absence of an indispensable party and failure to state a claim in the same category in establishing when they must be presented.3
We find some assistance in considering this issue in a recent Supreme Court case which reaffirms the inherent power of the court to consider certain legal issues as required by the interests of justice despite the failure of the parties to preserve them in a timely fashion. In City of Newport v. Fact Concerts, Inc., — U.S. —, 101 S.Ct. 2748, 69 L.Ed.2d 616 (1981), defendant city failed to follow the procedure required under Rule 51 of the Federal Rules of Civil Procedure to challenge the jury instructions regarding assessment of punitive damages. Nonetheless, the Supreme Court reached the merits of the issue, and stated: “Because the District Court reached and fully adjudicated the merits, and the Court of Appeals did not disagree with that adjudication, no interests in fair and effective trial administration advanced by Rule 51 would be served if we refused now to reach the merits ourselves.” Id. at 2754 (footnote omitted). The Court declined to limit the exception to Rule 51 to instances of “plain error”, and also noted the novelty and importance of the punitive damages issue involved in that case. The Court concluded: “In light of all these factors, we conclude that restricting our review to the plain error standard would serve neither to promote the interests of justice nor to advance efficient judicial administration.” Id. at 2754 (footnote omitted).
We believe that in unusual circumstances the courts of appeals also have inherent power to reach issues although not timely raised below. We have previously done so when the occasion warranted. See, e. g., Spicer v. Hilton, 618 F.2d 232, 239-41 (3d Cir. 1980); Allen v. Aytch, 535 F.2d 817, 819-21 (3d Cir. 1976); United States ex rel. Gittlemacker v. County of Philadelphia, 413 F.2d 84, 86-87 (3d Cir. 1969), cert. denied, 396 U.S. 1046, 90 S.Ct. 696, 24 L.Ed.2d 691 (1970).
Thus we proceed to consider whether this is an appropriate case in which to exercise our power in this regard. At the outset, it must be noted that the reasons given by defendants for their failure to raise the nonretroactivity issue are unpersuasive. Before the trial in this case commenced, other courts had already dealt with the question of the retroactivity of Elrod, and had decided that issue in favor of defendants’ position. See Ramey v. Harber, 589 F.2d 753 (4th Cir. 1978), cert. denied, 442 U.S. 910, 99 S.Ct. 2823, 61 L.Ed.2d 275 (1979); Litwhiler v. Hidlay, 429 F.Supp. 984 (M.D.Pa.1977). In fact, another judge in the district in which this case was pending had considered and filed a reported decision on this very issue. See Retail Clerks International v. Leonard, 450 F.Supp. 663 (E.D.*1362Pa.1978). Although that court found in favor of retroactivity, it should nonetheless have put defendants on notice of the possibility of raising that defense.4
Viewing the equities from Weaver’s perspective, defendants’ failure to raise the claim in a timely fashion obliged Weaver to expend the effort and expense of a trial and numerous post-trial motions which might have been avoided had the claim of nonretroactivity been presented to the district court and been decided favorably to defendants. On the other hand, we cannot ignore the somewhat ironic fact that because defendants failed to raise this claim, Weaver was enabled to succeed on the merits and as a result has been reemployed by defendants since November 1979, a job he otherwise would not have had.
In this case, therefore, it is not the individual interests of the litigants which impel us to exercise our power to reach the merits of the issue of the retroactivity of Elrod in this case. We are constrained to do so to vindicate one of the most fundamental jurisprudential concerns, evenhandedness in the administration of justice. The ultimate foundation of our system of law and order is the public’s peaceful acceptance of and acquiescence in judicial decisions. A bulwark of this public support is the public’s confidence that the courts strive to effect justice or at least substantial justice. Through a coincidence of timing, Weaver’s case, filed two years before Marino’s, appeared on the docket of this court at approximately the same time as Marino’s, and was therefore scheduled for argument at the same time. Both men had been in similar situations, having been hired by the same municipal entity at approximately the same time by the same Democratic administration and fired at approximately the same time when the Republican administration took over. We believe that the importance of evenhandedness in application of significant legal rulings overrides the necessity of following the rules of pleading in this unusual instance. We do not decide what the result would have been had Weaver’s case reached us before our decision in Marino or at some other time under some other circumstances.
We take this occasion to reiterate that all litigants must strictly observe the time requirements imposed by the Federal Rules of Civil Procedure; we limit our decision to the unusual situation at hand. Accordingly, we will reach the merits of the nonretroactivity defense and apply today’s ruling in Marino v. Bowers in this case as well. We will reverse the judgment of the district court which applied the Elrod decision retroactively, and direct that judgment be entered for defendants.5
Because we view defendants’ failure to raise the nonretroactivity issue in a timely fashion to have been without legal justification, we consider this to be an appropriate occasion to direct that defendants, notwithstanding their status as the ultimate prevailing parties, pay Weaver’s costs on this appeal. See Fed.R.App.Proc. 39(a) (costs of appeal are taxed against nonprevailing party unless court directs otherwise). Other courts have exercised their power to do so in similar situations. Dailey v. Payne, 22 Fed.R.Serv.2d 199, 201 (4th Cir. 1976); Thompson v. Sun Oil Co., 523 F.2d 647, 650 & n.4 (8th Cir. 1975); McEnteggart v. Cataldo, 451 F.2d 1109, 1112 (1st Cir. 1971), cert. denied, 408 U.S. 943, 92 S.Ct. 2878, 33 L.Ed.2d 767 (1972); Harris Truck Lines, Inc. v. Cherry Meat Packers, Inc., 303 F.2d 609, 612 (7th Cir.), vacated on other grounds, 371 U.S. 215, 83 S.Ct. 283, 9 L.Ed.2d 261 (1962). In addition, because defendants’ lack of diligence may have been responsible for unnecessary proceedings below, we will direct that they be required to recompense Weaver for all court costs, exclusive of counsel fee, incurred in the district court post-trial proceedings.
Accordingly, this case will be remanded for such further proceedings in accordance with this opinion.
*1363WEIS, Circuit Judge, with whom GIBBONS, Circuit Judge, joins, dissenting.For the reasons stated in my dissenting opinion in Marino v. Bowers, No. 80-1395, 657 F.2d 1363, decided this day, I would hold that Elrod v. Burns, 427 U.S. 347, 96 S.Ct. 2673, 49 L.Ed.2d 547 (1976), should be applied retroactively and, hence, would affirm the judgment of the district court. The “evenhandedness in the administration of justice” which the court applies in vacating Weaver’s judgment because recovery was also denied to Marino, demonstrates once again the difficulty with the concept of nonretroactivity. Weaver may well wonder why “evenhandedness” requires that he be treated the same as Marino, but unlike the successful plaintiffs in Elrod v. Burns, and indeed, unlike scores of other public employees who were discharged under similar circumstances just a few months later.
. For the period 1974-76, Weaver’s immediate supervisor was Marino, plaintiff in Marino v. Bowers.
. Although some language in Elrod v. Burns could have been construed as limiting its holding to non-policymaking officials, the Supreme Court subsequently ruled in Branti v. Finkel, 445 U.S. 507, 518, 100 S.Ct. 1287, 1295, 63 L.Ed.2d 574 (1980), that the “ultimate inquiry” is “whether the hiring authority can demonstrate that party affiliation is an appropriate requirement for the effective performance of the public office involved.”
. A possible basis for distinguishing between the two defenses in this regard is that the analysis which must be undertaken under Rule 19 includes consideration of the time and expense of a completed trial, if any. Provident Tradesmens Bank & Trust Co. v. Patterson, 390 U.S. 102, 111, 88 S.Ct. 733, 738, 19 L.Ed.2d 936 (1968).
. We cannot resist observing that throughout the post-trial proceedings defendants consistently sought to strictly apply various time requirements against Weaver.
. In light of our disposition of this case on the retroactivity issue, we need not reach defendants’ contentions with regard to their immunity.