Instead of returning hostile "500" errors to users, a small async worker query pattern makes fault-tolerant design the default.

In this talk, we'll talk about (1) how backend/database design impacts frontend experience, (2) the pervasive insanity of hand-rolling two-phase commit, (3) Postgres as a platform, (4) how async DB queues form a natural complement to Jira, and (5) how the tiny async query pattern works, and (6) how to tweak the query for optimized parallelism, prioritization, and sampling.