Comment 23 for bug 1931737

Hi all,

I've rebased this again, and am re-up'ing for eyeballs.

The two commits together provide protection (garuanteed by Postgres) against the deadlocks that concurrent symspell work can cause, and provide a way to delay symspell's lock-heavy work so that parallelization is effective. This should substantially remove the visible ingest-time performance effects of the symspell.

https://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/miker/lp1931737-dym-deadlock-plus-pingest-rebase