I have reviewed this patch and it still applies to master, compiles and passes the test suite.
I like the goal of the patch, making use of the already existing abbreviation machinery in more cases is something we should do and the patch itself looks clean.
I can also confirm the roughly 25% speedup in the best case (numerics which are all distinct) with no measurable slowdown in the worst case.
Given this speedup and the small size of the patch I think we should apply it. I will set this patch to "Ready for Commiter".