I can understand the sentiment that you should be prepared for even horrible bugs like this if you're running -CURRENT, but "if you can deal with crashing, you can deal with this" is completely wrong. Instability is a completely different beast from potentially generating easily crackable crypto keys.
I'm not sure what you're trying to argue for/against unless you are looking at things from a pure number theory/computer science research point of view? What would people do then? Create perfect flawless code that never malfunctions even as they write it? Code that never has bugs does not exist.
Perfect algorithms don't exist either.
-CURRENT is exactly the place for this because RNG bugs exist in reality.
If you read emaste's post above, this bug "potentially generates easily crackable crypto keys" for the safe to a monopoly game that holds monopoly money.
It's a development kernel.
Normally you could use a system like this on a trial basis and keep the results. For example, you might run some image processing on it. You'd want to make sure the output was good, but having done so you could then use those images for real-world tasks without worry.
But replace "image processing" with "generate a private key" and now you're deeply screwed.
Well, now we know. Don't generate crypto keys on pre-release OSes that you're going to use for anything important. That's reasonable. But I don't think many people would have thought that way before.
And of course pre-release OSes aren't the only place the danger lies. Debian had a similar problem that made it to releases and stayed undetected for a couple of years. The point is just that "silently break all your crypto" is a much worse vulnerability than mere crashes, or even corruption.
My point is that -CURRENT may be rather unstable at any given time, and is explicitly not supported. One should not be using it a production deployment where sensitive key material may exist.
In that regard the 'crashability' of the branch is indeed more significant than a random-numbers exploit: the crashability of the branch is a mnemonic for its general permissiveness towards unaudited (or not-sufficiently-audited) code. When you remember that your adversaries can inject code into the OS, that changes your threat modeling.
Of course, as you say, not-crashing back-doors can make it into releases, too. Open source's generic promise that more eyes will make bugs shallow is not a guarantee.