I’m reminded of a site I was called in to help rescue during the pandemic. It was a site that was getting a lot higher traffic (maybe 2-3x) than they were used to, a Rails app on Heroku. These guys were forced to upgrade to the highest postgres that Heroku offered - which was either $5k or $10k a month, I forget - for not that many concurrent users. Turns out that just hitting a random piece of content page (a GET) triggered so many writes that it was just overwhelming the DB when they got that much traffic. They were smart developers too, just nobody ever told them that a very cacheable GET on a resource shouldn’t have blocking activities other than what’s needed, or trigger any high-priority DB writes.
