My C is too awful at this point to get over my impostor syndrome and contribute to Postgres, but I've always been curious about the prospects of scaling Postgres higher by having different indexes on different replicas.
Lacking anything like this, we have to make the OLAP vs OLTP workflow splits sooner than we likely would have. Or at least, for companies that are growing organically. For VC backed companies this may be a difference of only 3-6 months.