supabase / etl / 30989127773

75%

push

github

Test dirty restart convergence (#949)

* Test dirty restart convergence

* Require replay after held destination response

* Abort apply worker when wait is cancelled

* Harden dirty restart test barriers

The dirty restart test waits for the table sync slot removal
before it generates load, so the crash always hits a state where
only the apply worker serves the users table. A comment explains
why drop is a crash and why the walsender poll gates the restart.

26 of 28 new or added lines in 1 file covered. (92.86%)

45866 of 61467 relevant lines covered (74.62%)

8773.04 hits per line