Performance metrics seem to include cached SQL queries in the query counts
In #43134 (closed) we ran into an issue where the SQL query thresholds were including cached queries. Quickly looking at our metrics code (
lib/gitlab/metrics) it seems we don't filter out cached queries there either. This means that for example the reported number of SQL queries may be far higher than the actual number of executed queries. We should make sure we filter out queries that are cached.