Does grouping as such then ever apply.
Suppose I have three cites: Doe 1985, Doe 2001, and Smith 1990.
If I sort by author I will naturally produce a group (Doe 1985, Doe 2001, Smith 1990). No “moving” is required. Collapsing can then apply simply to contiguous cites with the same names (year collapse) or years (year-suffix collapse).
But suppose I have a reverse sort by year. That gives me Doe 2001, Smith 1990, Doe 1985. Do I say “I have a sort so group” (Doe 2001, Doe 1985, Smith 1990), or do I say “don’t group because you are messing with my carefully constructed sequence”?
It seems to me arguable that one could
simply dispense with grouping, as such, altogether. If it happens, it happens. If it happens, then collapsing may remove redundant portions of any group which happens to be there. That’s all. It’s up to a style author to use sorting to create the groups, if that is what is intended (which, after all, is not that hard to do: a sort by name / year / year-suffix will do the job if you want aggressive grouping.
If that’s right one can simply remove all reference to grouping from the spec, and explain collapsing alone.