Currently 80K+ replication packets and 54K+ JSON incremental dumps in just two directories. It is impractical to browse these directories and even to keep on a single server.

These packets and dumps become obsolete after a few months for updating database mirrors and JSON full dumps. The only reason we keep them much longer is for debugging schema changes and issues that are not visible from the editing history.

A solution discussed with bitmap and I is to:

1. put those under separate directories following the schema sequence number (... 26/ 27/ 28/ ...) for the background task runner only but keep them in the same directory for FTP (so as not to break compatibility with any version of the server that would be needed to replay old changes).
2. keep only stuff for the current and the previous schemas on the same machine (which is currently trille experiencing disk usage issues).