@kornel Can we easily get the same graph weighted by usage stats (e.g. either number of crates depending on it) or limited to creates that had their first version more than, say, 6 months ago?
I expect with no evidence that a large part of the crates that depend on a new(ish) compiler version are ones that are also very new and virtually unused.
@robryk The "newly updated" chart excludes crates with 1 release only.
I think it's representative. There are crates like object and zeroize which end up in almost every project, and have high MSRV.
If you have 100 transitive deps, it doesn't matter that 99 of them are compatible.