If a shard node that is acting as a router for an internal transaction is killed or terminated while running a non-retryable transaction, the transaction would be left in-progress since nothing would aborted it. Such dangling transactions can cause chunk migrations to hang, for example, get stuck trying to enter the critical section. Given this, this workload also needs to set lowerTransactionLifetimeLimitSeconds like in internal_transactions_sharded_from_mongod_kill_sessions.js.