easy:
sorting with a btree index only scan
use count(*)
to implement aggregateless tables
medium:
better parsing https://github.com/pganalyze/queryparser/tree/master/extension
multiparameter aggregate functions (will affect tdigest, datasketches, allows implementing lots of built-in aggs)
use aggregate combining functions by default
better tracking of dependencies using postgres built-in features
benchmarks
good interface for settings
hard
outer joins
subqueries