However I
don't see why this couldn't generalize to any sized tuples. As long
as the tuples are all the same length, and the limit is one tuple
per key expression, then we don't get any combinatorial blowup
issues.
I think it's worthwhile to generalize for any sized tuples. I don't have any existing particular use case in mind, because BIP-44, BIP-84, etc. are fine with just using <0;1>, but there might be some upcoming standards in the future that will want to introduce more sub-paths.