Currently the implementation has a few things hard-coded (mostly due to speed improvments).
The logfiles considered are /var/log/syslog, /var/log/syslog.1, /var/log/syslog[2-31].gz - see:
https://git.proxmox.com/?p=pmg-log-...62995ed065b49d5039379bc485405aa;hb=HEAD#l2157
This means the maximal `rotate` value that makes sense is 32
speedwise I think it's a tradeoff between CPU speed, vs. disk-speed (if you keep weekly logs around each log will be ~7 times as large as for daily (more diskreads), if you rotate more often more logs will need to get gunzipped (more cpu-cycles)
pmg-log-tracker is built to only look at the beginning of the files and to only go through the ones that cover the provided timeframe - so I think the difference should not be too noticeable (if you have any concrete numbers in one direction or the other it'd be still very interesting to know)
all logfiles that fit the pattern from above are considered.
there are a few enhancement requests to make this all a bit more flexible (without sacrificing performance) e.g.:
https://bugzilla.proxmox.com/show_bug.cgi?id=3657
I hope this explains it!
