[TheBeast85]

What

Look up the last report in a FROM subquery and pass its id to report_severity

Why

task_severity passed the report lookup straight to report_severity, so a task that never produced a report called report_severity(NULL, ...). That returns NULL, but only after the whole severity query has been evaluated, including the report_counts lookup and the dynamic severity check. The cost is per task, so it shows up wherever severity is aggregated over many tasks.

With the subquery there are no rows when the task has no report, so report_severity is never called and the result is unchanged.

On an installation with 4507 tasks the aggregate goes from 2782 ms to 376 ms. The saving is proportional to the number of tasks that never ran; where almost none of them have a report it was 2774 ms to 9 ms.

Checklist

• Tests: severity distribution over all tasks identical before and after

[clawosiris]

Review result: I found no blocking correctness issue in the SQL rewrite.

I verified the exact current head (19a95738) against PostgreSQL 17. The new FROM subquery skips report_severity for tasks with no completed report (including missing tasks and tasks with only non-completed reports), while targetless-task behavior, latest completed report selection, legacy date ordering, and the overrides/min_qod arguments remain unchanged. The exact head also passes Debug and Release builds plus all 11 existing CTests.

There is a meaningful coverage gap: none of the existing tests executes the installed task_severity SQL, and this change is duplicated across four schema-version branches. I opened draft #3209 with a PostgreSQL-backed regression for all four definitions and tagged @TheBeast85 there. Its local validation passes all 12 CTests.

One integration caveat: #3070 is currently 219 commits behind main, and only the July automation checks ran on the PR. The commit applies cleanly to current main; a combined current-main simulation with #3209 also passes a Release build and all 14 CTests. I still recommend rebasing so the current hosted build/test matrix runs before merge.

Agent: GEA
Model: openai/gpt-5.6-sol
Thinking: xhigh
Bot-Owner: @llunved

[TheBeast85]

Thanks for the review and for the regression test in #3209! Rebased onto current main, no conflicts and no content change. Release build and all CTests pass, including the task severity test from #3209 applied on top.

[daniel-grnbn]

@TheBeast85 if that is OK with you, I'll remove the draft label from #3209 and merge it into this branch?

[TheBeast85]

@daniel-grnbn Yes, please go ahead and merge #3209 into this branch. Thanks for the tests!