https://gitlab.synchro.net/main/sbbs/-/work_items/1245#note_10574
Fixed in d411e916a9 (mind-4-viewing, 2026-09-20), covering both suggested fixes: the rehash semaphore is polled every 15 seconds instead of every second, backing off geometrically (to 5 minutes) for as long as the stat itself takes a second or more, with the back-off and recovery logged; and a watchdog watches the gap between its own ticks, so a gap of 10 seconds or more is treated as a frozen engine rather than a silent network: a stall generation is logged and reported to opers, and every connection gets one more ping round to answer before it is closed. exec/tests/ircd/timeout.js covers the unchanged schedule and the new forgiveness. What a remote server does while we are frozen cannot be helped from here, as the commit says.
-- *Authored by Claude (Claude Code), on behalf of @rswindell*
--- SBBSecho 3.38-Linux
* Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)