The DropSessionsCollection hook can timeout with "operation cancelled" errors when running in suites that also execute tests calling checkMetadataConsistency, due to DDL lock contention.

The current suites affected are: sharded_multi_stmt_txn_jscore_passthrough, multi_shard_multi_stmt_txn_stepdown_primary_jscore_passthrough, bulk_write_sharded_multi_stmt_txn_jscore_passthrough (and their variants)
Test triggering issue: check_metadata_consistency_basic.js

Root cause

1. DropSessionsCollection._sc_block_sessions_refresh() sets the failpoint preventSessionsCollectionSharding on all CSRS nodes
2. _sc_drop_collection() then attempts to drop config.system.sessions and call _flushRoutingTableCacheUpdatesWithWriteConcern
3. Concurrently, checkMetadataConsistency acquires DDL locks on shards via _shardsvrCheckMetadataConsistency
4. The drop operations conflict with CMC's DDL locks and fail with transient errors
   with_naive_retry retries for 100 seconds until timeout
5. The failpoint is never turned off because _sc_unblock_sessions_refresh() only runs after successful drop

We could for example ensure the failpoint is turned off in a finally block even if drop fails.