Thanks for sharing the gory details! Externally, the issue was not that intrusive on the livestream, I think your short term resource bump worked very well for buying time.
BTW, the burstable IOPS quota is a footgun on AWS too. Really appreciate the details and that we can all learn from it!
I can't help but think that "burstable disk I/O" is a marketing euphemism that has caused many headaches. Like, who _wouldn't_ want "burstable disk?" It sounds like a feature... Turns out, everyone; no one wants burstable disk.
Database outage on May 18, 2021 | Observable | Observable