History writes deliberately give up after a bounded 250 ms wait so they
never stall ComfyUI; the existing bounded-timeout test still proves that.
The concurrent replay tests assert that every peer succeeds, which is a
serialization and first-capture property, so they now raise the budget
to the shared 30 s busy timeout while the first writer is held open.
The budget is one named constant, EXECUTION_BUSY_TIMEOUT_MS, shared by
the transaction helper and the background writer.
Before: 20/20 'database is locked' with the held writer. After: 0/20.
The concurrent replay tests expected every peer to succeed, but each peer
only waits 250 ms for BEGIN IMMEDIATE by design. On a slow Windows runner
three queued commits exceeded that and a peer raised 'database is
locked'. Holding the first writer's transaction open for 500 ms makes the
same failure deterministic on any host (20/20 locally).
* fix(maintenance): keep titled and history-referenced prompts when pruning
prune_orphaned_prompts now also keeps prompts with a prompt_library_state
row (titled or explicit Library parents, whose state would cascade away)
and prompts that are the currently valid navigation hint of a recorded
run (execution_run_contexts positive/negative_prompt_id in the current
storage epoch whose exact raw snapshot text still matches). Stale hints
that history reads already ignore do not block pruning.
Fixes#237
* fix(maintenance): mirror unique-text hint rule and compute live hints once
The prune exempted a hinted prompt even when another prompt shared its
exact raw text, although history reads treat such a hint as dead. The
live-hint set now applies the reader's rule (current storage epoch, exact
raw text matching exactly one prompt, which is the hinted one) and lives
beside _prompt_hint in execution_contexts as LIVE_HINT_CTES.
The set is materialized once per statement and excluded with NOT IN, so
the prune no longer scans execution_run_contexts once per prompt. No
schema or index change.
Refs #237
Vito Sansevero