txn_being_applied_to_secondary_cannot_be_killed.js tests that the periodic job for aborting transactions running longer than the transactionLifetimeLimitSeconds parameter can't kill a transaction while it's being applied on a secondary. It does so by lowering the lifetime limit to 10 seconds, pausing transaction oplog application on a secondary, committing a cross-shard transaction, then sleeping for 30 seconds to give the periodic job enough time to run.

The transaction lifetime parameter is lowered on all nodes, so it's possible the transaction is killed (through the periodic job or the transaction coordinator's deadline) before it can be committed, failing the test. The test should be modified to handle this or at least make it less likely.