Recheck primary and replica dbs for lock_manager lwlock contention
As a follow-up to #2301 (comment 1391439437), recheck the main
db (and optionally other postgres dbs) for lock_manager
lwlock contention. We wanted to follow-up a few months after upgrading the VMs' machine-type and the Postgres major version.
In particular, as we introduce partitioning, there is a greater chance of hitting this bottleneck again.
If needed, we can add a query to postgres_exporter
to count the number of slowpath locks, for trending purposes.
Details of the analytical methods are in the linked issue.