HAND, Circuit Judge.

The plaintiff appeals from a judgment of Judge Ryan, summarily dismissing its amended complaint upon the merits. Included in the appeal are a number of other orders whose correctness it will not be necessary to consider, because we hold that the amended complaint was properly dismissed on the merits. The jurisdiction of the court was based upon diversity of citizenship, and in substance the complaint, including the exhibits which it annexed, together with the affidavits disclosed the following situation. On June 29, 1933, one, Koch, filed an application in the Patent Office for a “hearing aid apparatus”; on August 25, 1933, one, Nicholides, filed another application for a similar apparatus, and on November 11th of the same year the defendant Greibach filed a third application. On August 31, 1934, the Patent Office declared an interference between certain claims prepared by the defendant, Pi-neles, upon these applications. Koch assigned his invention to the plaintiff; Nicholides assigned his to the Sonotone Corporation, and Greibach gave an exclusive license to the Sonotone Corporation. In October, 1934, Koch moved to dissolve the interference as to Nicho-lides and Pineles consented. Hence, the situation after October of 1934 was that the interference was only between Koch and Greibach. On February 11, 1936 the examiner of interferences awarded priority to Greibach, and on October 5, 1936 the Board of Appeals of the Patent Office affirmed the examiner’s decision. Koch took an appeal to the United States Court of Customs and Patent Appeals and while it was pending the plaintiff and the Sonotone Corporation, on May 7, 1937, entered into a contract under which, in the event of an award of the invention to either party upon a final decision of the interference proceeding, the successful party would give to the other a non-exclusive license, for which the licensee would pay one dollar for every “bone receiver” sold. On the same day, Sonotone Corporation, for a gross sum paid, agreed to grant an exclusive license to the plaintiff under an existing patent, which, so far as appears, had nothing to do with the invention in question. On May 31, 1938, the United States Court of Customs and Patent Appeals affirmed the decision of the Board of Appeals and in due course Patent No. 2,127,468 issued on this decision. Koch v. Greibach, 96 F.2d 843, 25 C.C.P.A., Patents, 1168.

Two questions were decided in the interference proceeding: one, that Grei-bach’s invention was prior to Koch’s, the other, which is not raised on this appeal, that Greibach had not lost his right to a patent through delay. The decision of the first issue turned substantially altogether upon two physical exhibits, one of which, “No. 11,” bore the date “V.10.32,” and the other, “No. *133 14,” bore the date “9-4-32.” Koch claimed that the original years on both of these had been “33” and that Grei-bach had fraudulently changed them to “32.” On the interference proceeding Koch swore that one, Nowak, who had been in the employ of plaintiff until no later than March 1933, was familiar with “all the development work” in the plaintiff’s factory, and that every instrument discovered there was “always demonstrated” to him. Koch also swore that one, Benway, approached him in October or November, 1932, and came to the factory with Nowak. On his own behalf Greibach testified that in March, 1933, Benway called on him and explained to him that he was negotiating with people who promised to have “a bone conduction vibrator” ready in a short time, the designing of which Benway asked Greibach whether he would undertake. Greibach said he would, and Benway told him that the people with whom he was dealing — No-wak and Bing — were both Vice-Presidents of the plaintiff. Shortly thereafter, Benway, Nowak and Ring came to Greibach’s home and he told them that he could manufacture a “vibrator” without external contacts. Nowak at the time said that the plaintiff had something of the same kind. Sometime in March, 1933, Nowak and Ring discussed with Greibach the cost of developing some new instruments, but the three apparently decided nothing, for Nowak and Ring left, telling Greibach that they would see him later. After another interview Nowak disappeared. At no time had he discussed the design of the plaintiff’s vibrator with Grei-bach; he had merely spoken of his intention of starting a new “hearing aid company.” Thus Greibach denied that he had ever got from Nowak and Ring any knowledge of the invention, or had appropriated it, fraudulently ante-dating his exhibits, so as to ante-date his conceded interviews with Nowak and Ring.

The complaint alleged that in the summer of 1950 the plaintiff for the first time learned of facts which would have resulted in a decision in favor of Koch in the interference proceeding. These were set out in an affidavit of Nicho-lides, used to defeat a motion by the defendants for summary judgment and are the basis for the action at bar. We quote them in the margin. 1 The relief demanded was that the Greibach patent *134 be annulled, that the Sonotone Corporation refund all the license money paid by the plaintiff, that the Sonotone Corporation and Greibach be enjoined from claiming any rights under the patent, and certain other relief not necessary to set forth. In its brief on this appeal the plaintiff has disclaimed any effort to annul the patent, so that the remaining relief demanded is only for the annulment of the license agreements, a refund of the money paid and other incidental recovery. The action was commenced on October 24, 1950, originally against only the Sonotone Corporation; but on May 25, 1951, Greibach and Pineles were added as parties. An amended complaint was filed on January 8, 1952 and the cause came up on the day calendar on April 11, 1955; and was set for trial on April 25. Some time earlier the defendants had moved before Judge Weinfeld for summary judgment, which he had denied on January 29, 1951. On April 22nd the plaintiff’s trial counsel became sick and went to the hospital, and after several adjournments and interlocutory proceedings not necessary to describe, Judge Ryan summarily dismissed the complaint on May 23. It would be necessary to consider these interlocutory proceedings and say whether the dismissal without prejudice for failure to prosecute was justified, were it not that the judgment granted a summary dismissal on the merits, not only, (1), of so much of the complaint as sought to annul the judgment of the United States Court of Customs and Appeals (which, as we have said, the plaintiff in any case has withdrawn), but also, (2), of the license agreements; and, (3 and 4), of the prayer that the Sonotone Corporation refund all payments and pay damages “by reason of alleged acts of the defendants.”

The first question is whether Judge Ryan was so bound to follow Judge Weinfeld’s denial of the defendants’ motion for summary judgment that we must reverse his summary judgment on the merits without considering whether it was right: that is, whether the denial had become “the law of the case,” and must be accepted thereafter without re-examination. Is the doctrine described by that phrase an inflexible rule of law, or only a cautionary admonition to be applied when the occasion demands it? The order denying the motion for summary judgment was not appealable, and the consequence of holding that Judge Ryan was forbidden to reconsider it, would be that, if the judgment of dismissal because of failure to prosecute was wrong, yet the complaint should have been dismissed on the merits, the defendants would be compelled to suffer the loss of time and money involved in a trial that from the outset was unnecessary. No one will suggest that the first judge himself may not change his mind and overrule his own order, so that the basis of the doctrine can only be that there are reasons why the second judge may not do so that do not exist when the first does. We can think of only two such reasons: (1) the second judge should defer to the rule of the first as a matter of mu *135 tual respect between members of the same court; (2) if he does not so defer, the defeated party may shop about in the hope of finding a judge more favorably disposed. The first reason is clearly untenable; judicial sensibilities should play no part in the disposition of suitors’ rights. The second reason has indeed much to recommend it, and, as a matter of practice, has been universally regarded a sufficient reason for treating the first ruling as conclusive. It is, - .however, quite another question whether under all circumstances it makes the first ruling immune from reconsideration. As we view it, the question is in substance the same as that which arises when an appellate court upon a second appeal is faced with an earlier decision of its own, especially if the earlier decision happens to be that of a different panel of judges. The second panel has unquestioned power — ■ “jurisdiction”- — to deviate from the first, but as a matter of practice it rarely does so; and there is more reason to refrain than if two different actions are involved, when the only question is whether to follow a precedent. In any event on both occasions there is no imperative duty to follow the earlier ruling — only the desirability that suitors shall, so far as possible, have reliable guidance how to conduct their affairs. Since in the case at bar Judge Ryan had concluded to dismiss the complaint for lack of prosecution, the action was to come to an end, and the plaintiff would have to appeal. The order denying the motion for summary judgment was no longer of any value to it, unless it could reverse the dismissal for lack of prosecution. Thus there was every reason to present the merits of the complaint to the Court of Appeals against that possibility; and that was possible only in case it had before it a dismissal on the merits.

It is true that in Commercial Union of America v. Anglo-South American Bank, 2 Cir., 10 F.2d 937, we reversed a similar judgment for the sole reason that it overruled an earlier order of another judge denying a motion for the same relief; and it would not be frank to deny that we intended this doctrine to be without any exceptions. We have cited the decision more than once since it was made and have recognized its correctness, but we several times questioned the inflexibility of its statement. We applied it twice 2 in cases where the second judge had decided a point that was already pending undecided in a separate proceeding, taking the decision, so to say, out of the first judge’s lap. There can be no doubt of its propriety in those instances. On the other hand in Potts v. Village of Haverstraw, 2 Cir., 93 F.2d 506, 509 and in Munro v. Post, 2 Cir., 102 F.2d 686, 688, we declined to apply it, because the second order did not overrule anything actually decided in the first, although it was based upon an understanding of the law at war with that on which the first had stood. The distinction appears to us not to be relevant, if the theory is unflinching. In United States v. Steinberg, 2 Cir., 1938, 100 F.2d 124, 125, we said: “Whether there is any invariable and unyielding rule, rather than a practice, usually desirable, we might wish to reconsider, if it became necessary” — as it has become in the case at bar. That this is the only proper use of the doctrine appears from Messinger v. Anderson, 225 U.S. 436, 444, 32 S.Ct. 739, 740, 56 L.Ed. 1152, where the Supreme Court, through Holmes, J., spoke as follows: “In the absence of statute the phrase, ‘law of the case,’ as applied to the effect of previous orders on the later action of the court rendering them in the same case, merely expresses the practice of courts generally to refuse to reopen what has been decided, flot a limit on their power.” This was in substance repeated in United States v. United States Smelting, Refining & Min. Co., 339 U.S. 186, 198, 70 S.Ct. 537, 544, 94 L.Ed. 750: “The rule of the law of the case is a rule of practice, based upon sound policy that when an issue is once *136 litigated and decided, that should be an end of the matter.” It is true that in the two cases last cited the question was of the conclusiveness of a former ruling of an appellate court in a new appeal in the same case; but, as we have said, there is no relevant difference between that situation and the case at bar. So far therefore as our original opinion must be read as meaning that the question is not one of discretion, and that on no occasion may the second judge reconsider a ruling of the first judge, we now overrule it; and we come to a consideration of whether the dismissal on the merits was right.

Since the plaintiff has withdrawn any claim to have the patent annulled, the only remaining claim is that the license agreements should be invalidated because they were obtained by fraud. In the face of the allegations in the complaint that Greibach had in substance purloined the invention from Koch, and had obtained the patent by fraud, it would obviously have been improper to dismiss it summarily, were it not for the decision of the United States Court of Customs and Patent Appeals. If, however, that is an estoppel as to whether Greibach made the invention independently and before Koch, the dismissal was right, because the estoppel was a final answer to the claim of fraud. The District Court, sitting in a “diversity” case, of course had jurisdiction to set aside the license agreement for fraud, though it had none to annul the judgment of the Court of Customs and Patent Appeals; but questions may be raised as to what law should govern its effect as an estoppel. Should a state court apply the same test of fraud to the estoppel of a federal judgment based upon federal law, that it would apply to the judgment of a state court? Does it make any difference, if the challenged federal judgment was rendered in a “diversity” action based upon the law of the state? Should a federal court in an action based upon diversity give the same answers as a state court? 3 In the case at bar we need not answer these questions, if the judgment of the Court of Customs and Patent Appeals would be an estoppel, whether the state or federal test be used. It is certainly true that the New York test is not more liberal than the federal test. 4

The decisions of the Supreme Court in Hazel-Atlas Glass Co. v. Hartford-Empire Co., 322 U.S. 238, 64 S.Ct. 997, 88 L.Ed. 1250, and its companion appeals, Shawkee Manufacturing Co. v. Hartford-Empire Co., 322 U.S. 271, 64 S.Ct. 1014, 88 L.Ed. 1269, may be taken as laying down as liberal a test as any other federal decisions and they are indeed the plaintiff’s chief reliance. It is important to state with some detail the facts in these cases, for the scope of the decisions has not always been understood. They came up upon petitions of the Hazel and Shawkee companies to the Court of Appeals for the Third Circuit, 137 F.2d 764, to be allowed to file bills of review of judgments of the District Court, which had been entered upon mandates of the Court of Appeals reversing judgments for the Hazel and the Shawkee companies as infringers of a patent, and directing judgments in favor of the Hartford company. That company upon the appeal from the judgments dismissing its complaints had laid special stress upon a published article, commending the invention in suit as a wide step forward in the art, although the article had not been actually put in evidence at the trial. In fact it was a fabrication and a forgery concocted by the Hartford company, and so both the Hazel and the Shawkee companies had been told during, or after the trial. In *137 deed, on the appeal in the Shawkee ease one of its lawyers had told one of the circuit judges that the article was spurious. Nevertheless, in its opinion reversing the District Court’s dismissal of both complaints, the Court of Appeals had particularly relied upon the fabrication, “quoting copiously from the article to show that ‘labor organizations of practical workmen recognized’ the ‘new and differentiating elements’ of the ‘gob feeding’ patent.” [322 U.S. 238, 64 S. Ct. 999.] After the District Court had entered judgment for the Hartford company on the mandate of the Court of Appeals, the Hazel and Shawkee companies upon investigation learned the details of the fraud, and it appeared without dispute that the Hartford company had bribed the writer of the article to fabricate and publish it in a trade-union paper. On these facts the Supreme Court reversed the Court of Appeals which had denied the petitions for leave to file bills of review, and dismissed the complaint. The Court held that, despite the fact that the validity of the patent had been one of the issues determined in the infringement actions, that the Hazel and Shawkee had had notice that the article might be a fabrication, and that the term had long since expired in which the judgments had been issued, the undisputed fraud had been so gross as, not only to excuse, but to demand, the intervention of a court of equity.

It is plain that the Court did not mean to annul the “salutary general rule * * * that in most instances society is best served by putting an end to litigation after a case has been tried and judgment entered.” 322 U.S. at page 244, 64 S.Ct. at page 1000. It did not decide that the unsuccessful party to an action may always reopen a judgment upon producing evidence of fraud in procuring it, even though the evidence was inaccessible at the trial; especially it did not so decide when the trial had turned upon the issue of fraud and the new evidence was another version of what had originally appeared. On the other hand, it did mean to abandon the distinction made in United States v. Throckmorton, 98 U.S. 61, 25 L.Ed. 93, between “extrinsic” and “intrinsic” fraud, as the Restatement of Judgments had already done. 5 Indeed,, it is impossible to draw any line based upon that distinction; for all fraud is designed to prevent, and, when successful, does prevent, the unsuccessful party from proving the facts that determine his rights. The fact that he does not at first succeed in unmasking the machinations of his adversary is no reason for denying him relief after he has finally succeeded in penetrating the complete disguise. And yet it is obvious that there must be a limit, else the mere assertion of relevant evidence, though it was inaccessible at the trial, will be enough to upset any judgment. From this dilemma there is no escape unless in each case we balance the conflicting interests and make a decision ad hoc; and that is as we understand the decisions of the Supreme Court that we have cited. Hence we must consider the occasion at bar in its concrete details and without the help of any controlling general postulate.

The story told by Nicholides which is the basis of the present action is that Nowak, shortly after leaving the plaintiff’s employ, came to Lieber, the president of the Sonotone Corporation, and Pineles, its attorney, in May, 1933, and showed him a model of Koch’s invention. Lieber told Pineles to draw specifications for a patent on a “bone conduction receiver” that Nicholides had already designed, of which the “disclosure was dated March, 1933.” Pineles did so, and filed the application on August 25, 1933. So far, there is no suggestion that the parties meant any fraud or fabrication; it was to be a straight contest between the two disclosures. However, Pineles found out later that Koch’s date was before March, 1933, and *138 told Lieber that Nicholides could not succeed. Thereupon the two decided “to make it appear that Greibach had invented the same bone conduction receiver * * * at an earlier date than the date claimed by Koch.” How long after Nicholides filed his application Pineles learned that Koch did antedate his invention does not appear, but the interference was declared on August 31, 1934, and Nicholides did not withdraw from it until October of that year. Moreover, Greibach had filed his application on November 11, 1933, and obviously Pineles must before that have prepared the specifications and set on Greibach to forge the marks on Exhibits “No. 11” and “No. 14.” Lieber .and Pineles had therefore not only con■cocted the fraud between August 25, 1933 and November 11, but during the interval Greibach, learning from Pineles of Koch’s apparatus, had fabricated the exhibits, perhaps separating them by four months to add plausibility to his testimony. Lieber had discharged Grei-bach in March, 1932 and Greibach was bitterly resentful, though Pineles remained on friendly terms with him and saw him not infrequently. Even though we suppose that sometime in September or October of 1933 Pineles learned that Nicholides’s application was too late, it remains unexplained (1) why they did not withdraw it for a year thereafter; (2) why Pineles, sometime before August 31, 1934, drew up interference claims which should read upon it, as well as upon Greibach and Koch; and (3) why he withdrew Nicholides from the interference a month later. Certainly it rested upon the plaintiff to make some explanation of such a hiatus in its evidence.

Furthermore, we cannot see what testimony Nicholides could give on a new trial, if there was one, that would be at once competent and discredit the evidence on which the interference proceeding was decided. Lieber is dead and did not testify at trial. Moreover, nothing that he said to Nicholi<3es was .in execution of the putative conspiracy, for the interview between him and Lie-ber and Pineles ended in Lieber’s asking Pineles “what could we do so that we could get priority of invention,” and in Pineles’ answer that he would talk to him later. It was “several days later” that Lieber told Nicholides “that Pineles would bring Greibach into the picture in such a manner as to predate Koch” etc.; and that was not in execution of the conspiracy. Thus, we do not see that any new evidence could be adduced; if so, all that could happen upon a new trial would be an added opportunity to cross-examine Greibach and Pineles. Finally, even if it be supposed that when Nicholides withdrew his application in October, 1934, he had become privy to the conspiracy, that would not make his earlier hearsay testimony competent. But suppose it did; it would then appear that he had willingly gone into a scheme with two rogues who were seeking to defraud the plaintiff and the court, and to steal Koch’s invention. We cannot agree that such testimony discovered after fifteen years can be the basis for reopening a judgment, especially a judgment that, turning as it did upon a charge of fraud, was likely to put the unsuccessful party on the alert for all possible support of that position. Nor are we met here with the objection that we must be jealous of validating any patent that is challenged, because — quite aside from the plaintiff’s disclaimer of any intention to annul this patent — if it has not already expired, little of its life can be left, and the public interest, which was a moving factor in Judge Weinfeld’s decision, by now has at best become negligible. We do not hesitate to hold that the considerations supporting the doctrine of res judicata should prevail.

The paragraphs numbered I and II of the judgment, which are concerned with the dismissal of the action for failure to prosecute, will be vacated, as they become moot with the affirmance of the paragraphs numbered IV, V, VI and VII. . The paragraph numbered III is vacáted because the plaintiff on this ap *139 peal withdrew the corresponding prayer for relief. The paragraphs numbered IV, V, VI and VII are affirmed.

Judgment will be entered in compliance with the foregoing direction.

Notes

. “What actually happened was that Nowak came to Sonotone Corporation some time during the latter part of May 1933 and showed to Dr. Lieber, Pineles and myself a working model of Koch’s bone conduction receiver and told us that it would be on the market in a month. At that time I made a sketch of Koch’s model. A photostatic copy of my sketch is annexed hereto as Exhibit T.’

“I know the date Nowak did this because immediately after his visit we got to work to produce a bone conduction receiver based on the same principle as Koch’s, patterned after my disclosure of March 30, 1933. The first thing I did within a day or two after Nowak’s visit was to make rough sketches of such proposed receiver. One of these, dated May 25, 1933, fixes the date of Nowak’s visit. A photostatic copy thereof is annexed hereto as Exhibit ‘2.’

“After Nowak left, Dr. Lieber accused mo, in Pineles’ presence, of being asleep at the switch in permitting Dictograph to develop a more practical bone con-duetion receiver before Sonotone. I told Dr. Lieber that I had been working on the development of a practical bone conduction receiver, and I showed him a disclosure which I had made to him. This disclosure was dated March 30, 1933. A photostatic copy of it is annexed as Exhibit ‘O’ to Franchot’s affidavit.

“Dr. Lieber then directed Pineles to prepare and file a patent application in my name, based upon this disclosure. Pineles did so.

“Pineles discovered and told Dr. Lieber in my presence that the date when Koch> claimed he made his invention was earlier than the date of my disclosure, and, therefore, in all likelihood, Koch wouldi obtain priority of invention in the Patent Office and Dictograph would obtain the patent on the bone conduction receiver and not Sonotone.

“Dr. Lieber then said to Pineles in my presence: ‘since we had a sketch of Koch’s model, which Nowak had shown to us, and my disclosure, what could we do so that we could get priority of in *134 vention?’ Pineles stated that he would discuss it with him later.

“Several days later, Dr. Lieber told me that Pineles would bring Greibach into the picture in such a manner as to predate Koch; that Pineles would set things up with Greibach so as to make it appear that Greibach had invented the same bone conduction receiver which was the subject of Koch’s invention at an earlier date than the date claimed by Koch, so that Sonotone could get priority of invention.

“It is to be noted that the exact invention of Koch for a bone conduction receiver was the subject matter of Grei-bach’s patent application, as well as my patent application.

“Pineles then prepared and filed a patent application in Greibach’s name.

“My application was withdrawn, since the date of my disclosure was later than the date claimed by Koch. Dr. Lieber had me sign a waiver, a photostatic copy of which is annexed hereto as Exhibit ‘3,’ in which I agreed not to contest the interference proceeding and disclaimed the broad claims of principle and application of principle involved in the counts in the interference proceeding.”

. In re Hines, 2 Cir., 88 F.2d 423; Application of Tracy, 2 Cir., 106 F.2d 96.

. Moore’s Federal Practice, par. 60.37(2), pp. 626-630.

. Crouse v. McVickar, 207 N.Y. 213, 100 N.E. 697, 45 L.R.A.,N.S., 1159; Jacobowitz v. Herson, 268 N.Y. 130, 197 N.E. 169, 99 A.L.R. 1198; Burbrooke Mfg. Co. v. St. George Textile Corp., 283 App. Div. 640, 129 N.Y.S.2d 588; decides no more than that the estoppel of a judgment as to one element of a conspiracy is not necessarily a bar to the claim.

. Restatement of Judgments, § 118, Comment b.