That's one of the things I'm wondering, if I change the prefix, will it need to traverse less data? I have a feeling that `clean_old_session_data` is traversing too much data. It should only look at the keys where prefix is KEY_PREFIX (find_by_prefix(cache_session::KEY_PREFIX)) and that is sess_ by default, but I don't have any keys with sess_ prefix in my redis.
Anyway, we'll create 2 MUC instances, one for app cache and another for MUC sessions and set different prefixes for all caches just to see if it fixed the issue.
As multiple instances of Redis go, we would like to keep it single for the time being and we wouldn't like to put anything on shared Moodledata as performance is useless based on our experience with NFS for sessions data.
thanks for ideas.