Some time ago, I posted this in Deprecating UnwindSafety PR:
I feel like the only goal of unwind safety (being an useful lint) hasn't really been achieved, because there's no consistency in implementing UnwindSafe and RefUnwindSafe even in std (!). Scanning through whole std and fixing it seems like a lot of effort for little gains. Also, contributors have to keep in mind that any future code modifications may invalidate type's unwind safety.
In my opinion, there's no place for half-baked features like this one in std.
it provoked some minor discussion and died, but I feel this topic shouldn't be forgotten without strong words on its state. Especially now, with the new edition on the horizon.
What's your feeling?