Not very necessarily.
D76746 touched the upper bound. The point of the upper bound is actually quite weak: it is used as an assert in MCFixup::create to reject certain apparently invalid relocations.
I want to keep the code as-is to be consistent with changes I made to ARM/AArch64/X86/PowerPC.
I think this looks good, but I definitely feel out of my depth in evaluating it, not really knowing this part of the backend and any invariants that need to be preserved. For this reason I'm unwilling to give an explicit approval/rejection.
I do think the testing seems adequate, and this is a useful feature that we should enable.