Yes, it is an inductive property. Sadly, not a very powerful one.
> On the benefit side: proving that libraries do not undermine Rust's safety guarantees is important, and the importance grows with the popularity of the library.
It is only important if you take it as an axiom that all programs must be proven to be memory-safe. That may be an axiom accepted by some Rust fans, but it is obviously not an axiom of software. What software needs are methods that reduce dangerous bugs (where "dangerous" means different things for different programs) in the most cost-effective way. It may well be the case that, say, soundly eliminating 95% of bugs caused by memory unsafety and using the rest of the budget on unsound methods is more effective than spending it on eliminating the last 5% of those particular bugs. Software's needs do not currently justify complete sound memory safety at all costs, so the question of cost is still the most relevant one. There is no justification to focus on one cause of bugs just because it can be soundly eliminated.
> And of course those guarantees aren't just "memory safety" but data-race-freedom and the general "mutable-xor-shared" property --- the importance of which is the subject of the OP here!
Neither of these are worthwhile enough to move the needle. This is precisely what I mean by the myopic focus on what can be done as opposed to what needs to be done. If the gap is 10 orders of magnitude, narrowing it by 1 doesn't have a big impact, unless it were completely free, which it isn't.