Where's the violation?

In this case Red Hat is both the distributor and the provider of support services. If they were not actually distributing GPL'd code, but simply offering support for people who happened to have it, there would be no issue (and there would be no SRPMs to make public or private). Red Hat is (voluntarily) bound by the GPL because that is the only thing that allows them to be a distributor; without it they would just be straight-up violating copyright.

(We should also be clear that "support services" includes providing software! The reason that Rocky is going for "bug-for-bug compatibility" with RHEL is that there are actual third parties who build software to run on top of RHEL and the exact behavior of RHEL is important. The whole commercial point of making it difficult or FUDdy to get the exact behavior of RHEL is that people will pay for the exact behavior of RHEL.)

> The other thing to bear in mind is the law believes in "hats". Never mind that the same physical person can be all three of the copyright holder, the distributor, and the support service - as far as the law is concerned they are three different entities because they have three different "hats".

At least in the US (the jurisdiction of Red Hat/IBM, Conservancy, the authors of the GPL, and many authors of the Linux kernel, which is probably the most interesting Red-Hat-redistributed software where Red Hat doesn't hold the full copyright), this isn't true.

There's a case going through the US courts right now, Davitashvili v. GrubHub, a class-action suit where customers of local restaurants are suing food delivery apps because the app agreements say that in-person takeout and delivery must be set at the same price, and they say that as takeout customers, they're harmed by this and this is illegal under antitrust law. But it turns out that most of the plaintiffs have signed up for these food delivery apps and clicked through binding arbitration agreements, and the defendants are trying to get the case moved into arbitration. The trouble here is that the plaintiffs are clearly acting with different "hats." They're not complaining about the service they got through the app, they're complaining that as takeout customers, the app that they happened to sign up with, long ago, is unfairly raising prices on them. Should they be required to settle this via arbitration? If the law really believed in "hats," the answer would be clearly no: the contract would only bind them as an app user. But that argument isn't being made at all. The argument is over the exact terms of the arbitration clause, what sorts of disputes it covers, and whether "infinite arbitration clauses" are legal under arbitration law. (The judge is saying no; it's being appealed, and the Chamber of Commerce has written an insistent amicus brief that contracts that broadly bind customers are very important to the economy. So they clearly don't believe in "hats" either, and as the representative of companies, they would have very good reason to believe in it!) Neither the plaintiffs nor the judge seem to be under the impression that there's a non-arbitration-specific argument that the app contract/license doesn't bind the plaintiffs when they're not using the app.

So I'm assuming you mean some other legal system - can you clarify which legal system you're talking about and point to some reference about this concept of "hats" in that system?