Hi, following up on Benoit’s mslab observation: we can confirm Assertion 'MSL_GET_STATE(ph) == slh_partial' failed at lib/mslab.c:789 as a recurring crash, not a one-off. Six+ occurrences between Sep 6 and Nov 3, 2026, on two identically configured edge routers, FreeBSD 15.1-RELEASE amd64, on BIRD 3.3.1 and 3.3.2. lib/mslab.c is unchanged between v3.3.1 and v3.3.3 apart from POISON constant renames, so probably all current 3.3.x should be affected and Benoit saw it on 3.4.0. The crash is always SIGABRT in the main thread while a worker thread is concurrently executing BIRD code consistent with a race between msl_cleanup_partial_heads() and concurrent mslab operations. The process runs with 3 threads total, so high parallelism is not required. Our routers run 5 BGP sessions, 3 of them with full tables; identically configured 3.2.x routers (no mslab) never crash. Unlike the rt-table.c:1562 cases discussed earlier in this thread, none of those conditions apply to our setup: we do not use “import keep filtered”, our iBGP sessions to the route reflector are direct (single-hop), and next hops are on connected subnets with no recursive resolution, the RR sets itself as next hop. The mslab.c:789 crash occurs regardless. BR Kalina Rakovska AS50837