Per our tuple statistics (https://performance.gitlab.net/dashboard/db/postgres-tuple-statistics?refresh=1m&orgId=1) we seem to be updating `projects` quite a lot over time:  This in turn also affects the number of dead tuples over time:  Looking at the last 90 days of data we can see the number of dead tuples per minute gradually increase:  I'm worried that some process is performing unnecessary updates on the `projects` table. We should reduce this value back to the old value of <= 5k dead tuples per minute, instead of 17k. cc @tiagonbotelho