This is pure speculation of my part, and I still use a ton of snappy, but Klaus Post's S2 and
now the next generation MinLZ would appear to have surpassed snappy in
performance while providing a backwards compatible migration path.
Old data snappy compressed can be read, and new data can compressed and decompressed faster.
So from a technical standpoint, there's not much call for snappy in new code, unless
you deliberately want to go slower than what is now possible. That is a rare want.
Links:
https://gist.github.com/klauspost/a25b66198cdbdf7b5b224f670c894ed5