Avoid fast spin storage recruitment on storage server with bad disk

If a host has a bad disk that prevents the kv-store files from being opened (e.g. with file_not_found errors), then the storage server can repeatedly die and be re-recruited in a tight loop. We should avoid doing this so quickly one after the other if it’s repeatedly failing.

Start by tracing storage-server recruitment and the kv-store file-open failure path described in the issue. Review how repeated failures trigger re-recruitment and determine how to prevent a tight loop. Done means a storage server with an unusable disk does not repeatedly die and get recruited in rapid succession.

Written by the indexing model from the issue text.