Which licenses, precisely, are you describing as "gratuitous"? Consideration is, after all, not a hard thing to find.
In the case of software using copylefted dependencies, the ability to use 3rd-party similarly licensed code is consideration for release the license. In the case of software under more permissive licenses, there's an argument to be made that public assistance in the development of same (bug reports, community support assistance, etc) acts as consideration for the license. If a single peppercorn is sufficient to establish compensation under common law, surely a well-researched bug report is worth more.
You ask for an example of a case when a "gratuitous" license (a term implying that absolutely no consideration is given, which I deny is the case in the situations given here) was not allowed to be withdrawn. Frankly, I'm not familiar with a single instance in which an OSI-approved license has been withdrawn in a US jurisdiction with respect to previously released codebases -- and were this a feasible thing, we'd have seen Oracle, SCO and others doing no end of it (particularly in the time period in which Microsoft was willing to spend money on convincing the world that using open source software in business was high-risk, and certainly had the funds to buy companies which owned copyright to the codebases of major OSS infrastructure, either directly or by proxy).
I'd be curious to hear about a case of revocation of an OSI-approved license being held valid in a US court, should such exist -- and suspect that, if one did make it to appeals, we'd be seeing the OSI and their friends weighing in as amici; it'd certainly be an interesting read.