Let me argue in favour of learned indexes by spamming the PGM-index, the (AFAIK) only learned data structure with formal worst-case query time guarantees.
i found this to be quite instructive as an overview of various hashing techniques. check it out :)
Cool! The overall series looks useful, as I’m implementing a key-value store too.
Of hashing techniques, the only fancy one I’ve tried (in a prior project) is Robin Hood, which I found worked well, increasing both speed and maximum load.
I found that bidirectional linear probing worked better than Robin Hood: https://github.com/senderista/hashtable-benchmarks