Seeing 'COLLECTION_MAP_REFRESH_IN_PROGRESS' in steady state conditions

Scenario:

1. Two node cluster, Couchbase bucket (replica=1, scope=3, collections=2 per scope)
2. Load each collection with 100 docs initially into all collections
3. Open a separate client for testing sync_write outcome
4. Stop persistence on one of the node using cbepctl command
5. Load single doc with durability=MAJORITY_AND_PERSIST_TO_ACTIVE on 5 collections in parallel with timeout=60sec, such that the op won't timeout
6. Using the client opened during step#3 try the same doc_op on same doc_key to validate  SYNC_WRITE_IN_PROGRESS outcome

Observation:

Seeing COLLECTION_MAP_REFRESH_IN_PROGRESS during one of the five operations. All other operations results in valid KV_SYNC_WRITE_IN_PROGRESS.

Expected behavior:

Should not see MAP_REFRESH_IN_PROGRESS, since the client has been already opened well before the doc_ops and the simulated error here is only STOP_PERSISTENCE(which has no direct effect on client)

Cbcollect logs:
https://cb-engineering.s3.amazonaws.com/c_map_refresh_in_progress/collectinfo-2021-03-23T120441-ns_1%40172.23.105.155.zip 
https://cb-engineering.s3.amazonaws.com/c_map_refresh_in_progress/collectinfo-2021-03-23T120441-ns_1%40172.23.105.211.zip
Attaching test run logs test.log and pcap col_map_refresh.pcap

QE test:

epengine.collection_crud_negative.CollectionDurabilityTests:

  test_sync_write_in_progress,nodes_init=2,replicas=1,num_items=100,sdk_timeout=60,doc_ops=create;create,bucket_spec=single_bucket.three_scopes_two_collections