>Do you think that it is important to do so? Are you experiencing problems which you believe are due to missing/out of date >activity statistics? Or is this more for curiosity?
I am making specifications for a tool which captures query plan statistics. I wanted its behavior to be the same as statistics collector for everything except in the kind of statistics being collected. This is the same premise used by pg_stat_statements. Hence the question wrt PITR recovery.
In the case of my tool there could be as case that user wants to see the information like 5 slowest queries after archive recovery. For this i would need to build in logic to ensure the stats file read in during recovery is in sync with that timeline. It sounds complicated so was wondering how statistics collector handled it.
Seems to me that the stats file is ignored in archive recovery mode,and deleted. This is consistent with its crash recovery behavior as well.