Just worked through a small problem with a friend who had sporadic issues with the wordpress “redis object cache” plugin. “socket read errors” were not uncommon, and sometimes “script already running” messages from redis itself.
I didn’t quickly get a handle on what the plugin is trying to cache, or how, but my sense is that it was bundling a lot of data in one request. The redis server in question was also set up to persist to disk. This shouldn’t be something they need. The source of the data is already in wordpress; the ‘caching’ should only be being used to speed things up, but not fail if the data is missing (redis gets reset or the value is evicted, for example).
Disabling persistence altogether (removing all ‘save’ references in the redis.conf) *seems* to have resolved the issue.