tl;dr:
Database did not have correct indexes.
Database is not scalable (serverless).
IOPS were hard limited by AWS plan.
No performance metrics/alerts in place to easily pin-point the issue.
Infrastructure resources are shared by all users (one user going viral means that all the other users are affected).
Article was somewhat confusing for me, as it didn't clearly state the problem (bottleneck), solution or architecture.
