Isn't the key part that these older systems have documented bugs?

Or, to put it another way, if there's a wasp in the room (and there always is), I'd want to know where it is.

That doesn't end up being the case for a number of reasons. Firstly, no one is actually able to account of all of these known issues a priori. I don't like calling things impossible, but writing safe C that avoids any compiler bugs is probably best labeled as that.

Secondly, vendors make modifications during their release process, which introduces new (and fun!) bugs. You're not really avoiding hidden wasps, just labeling some of them. If you simply moved to a newer compiler, you wouldn't have to avoid them, they'd mostly be gone (or at worst, labeled).

Are the newer compilers truly that much better? I've been working in tech since the 90's, and I can't say that for the tools I've used I've noticed any great improvement in overall quality- bugs get swatted, new ones created in what feels like a constant measure. I am assuming that many optimizations are turned off regardless, due to wanting to keep the resulting assembly as predictable as possible, but I do not work in the embedded space, so this is perhaps a naive question.

I think the idea is that you don't want a whole wasp nest. Just a bunch of stray wasps.