A case for GIST supporting ORDER BY

Some time ago I’ve provided some details with the issues we face when trying to use GIST and partitioning at the same time in the postgresql-general mailing list:

We decided to go with the solution to partition our table by:

RANGE (‘2100-01-01' <-> operation_date).

While it (somewhat) solves partition pruning issues described above there is another problem:

It is impossible to create a unique constraint on the partitioned table.

So now we cannot use INSERT … ON CONFLICT (…) DO UPDATE

My question to hackers:

Would it be feasible to implement ORDER BY column GIST index (only) scan for types with total order and sensible greatest and least values?

Thanks,

Michal