Sometime back I did a Golang based memcache server implementation with BoltDB (experimented with Pebble as well). I dropped that implementation out of fear that nobody will be interested in using it, have same question here I guess. Why would I drop the existing battle tested solution for something that might still have non-memory bugs.