A long time bottleneck in performance is leveldb and berkeleydb, so investigating something like what bitcoin is doing
bitcoin/bitcoin#19077 would help begin our long term goal of changing dbs in the whole code base.
Preferably moving to something like upscaledb or some other graph db, with some work, will allow for faster queries and allow for indexing relationships between transactions far more efficiently. The end goal is to update the local storage of the blockchain as well. Also, seeing as how we are not constrained by bitcoins "no hard fork" mantra this should be pretty easy to accomplish, I doubt that this change would need a hard fork.
https://upscaledb.com/