Problem being that sets aren't consistently ordered and conversion to a tuple can result in an exponential (specifically, factorial) explosion in the number of possible keys associated with a single set. Nor can you sort all objects. Safe conversion of sets to tuples for use as keys is possible but the only technique I know requires an auxiliary store of objects (mapping objects to the order in which they were first observed), which doesn't parallelize well.
tuple(sorted(s)) and if you can't even sort the values, they're probably not hashable. I get that this involves a copy, but so does frozenset, and you can cross that bridge in various ways if it's ever a problem.
I'm not saying the problem with tuple doesn't exist, but that there doesn't need to be a built-in way to deal with it. If for some unfortunate reason you've got a mixed-type set that you also want to use as a dict key, you can write a helper.
