The WASM JS engine processes the 18 MB MapReduce workload in mr_bigobject.js ~10× slower than legacy SpiderMonkey on arm64 CI hosts vs running the test locally (~131s vs ~13s locally, ~1s legacy). This gives the  DoReconfigInBackground hook enough time to fire and interrupt the WASM execution with wasm trap: interrupt.  

Workaround: mr_bigobject.js is excluded from replica_sets_reconfig_jscore_passthrough until WASM MapReduce performance is acceptable on arm64.

Root cause: all BSON data must be copied twice across the WIT boundary in each direction. For an 18 MB MapReduce payload this amounts to ~100+ MB of memcpy that the legacy path avoids entirely.