honest question because i've gone back and forth on this myself.

when sentry fires do you actually reproduce it locally as a failing test before touching anything, or do you just read the trace, understand what broke and push the fix?

i always end up spending like 30-45 mins just getting the repro right. reconstructing the state, getting deps working in the test, running it, realizing the inputs are slightly off, running it again. by the time it actually reproduces i've lost the whole debugging flow.

got annoyed enough that i started building something to automate it. grabs the frame locals from sentry, generates a pytest, runs it in docker against your branch. still figuring out if this is actually useful to other people or just my own problem.

how long does it take you to write a repro test from a sentry trace? do you even bother or just push and monitor? has skipping it ever come back to bite you?