There are reasons to desire invoking the same hook twice in two groups of the same stage - namely, if there is any I/O that can be started in one group and joined/Wait'ed in the 2nd - then the I/O time can overlap/be parallel with the CPU time of the first group.

While there are techniques such as putting a counter into an entry of the ModuleInvocationContext.ModuleContext, it seems like a missed opportunity to just use the configured hook_impl_code - if this were passed in the ModuleInvocationContext then the module could easily differentiate which of the two invocations were being invoked definitively.