DGM: Bucket resident ratio is calculated wrongly when moving from dgm to non_dgm scenario

Build: 6.5.0-3840-enterprise

Scenario:

1. 3 node cluster, Couchbase bucket with ram quota=100MB (replica=1)
2. Initially load 10k docs
3. Load docs into bucket until dgm of 80% is reached.
4. Remove inserted docs starting from doc_number 10K. (Docs inserted in step-3)
5. In parallel add 10K of new docs and update docs from 0 - 10K (docs from step-2)

Observation:

After moving out of DGM scenario also, the resident ratio is given as 99.4% even though only 1/3 of the RAM size is only utilized.

Not able to reproduce the issue while running the same case again (This time it said 100% correctly).

Expected behavior:

Resident ratio is expected to be only 100% after deleting all the newly inserted docs.

Bucket and cluster status

TAF testcase:

bucket_param_tests.dgm.Bucket_DGM_Tests.test_dgm_to_non_dgm,nodes_init=3,num_items=10000,doc_size=1024,active_resident_threshold=80,replicas=1,sdk_timeout=60,bucket_size=100