Data in few datasets is not getting flushed when bucket on which dataset is created is flushed

Steps to reproduce -
1. Create a cluster with 2 KV and 2 CBAS nodes.
2. Create a KV bucket with 1GB memory allocated to it.
3. Create 10 scopes and 2 collections in each scope.
4. Load data into collections such that the bucket is in 40% DGM.
5. Now create datasets on all the collections of the bucket.
6. Wait for ingestion to be completed into datasets.
7. Flush the bucket.
8. It is observed that for a few datasets the data is not deleted from even though the collection on which it was created is empty.

In the logs look for dataset `6jHjE6`.nZj6.NxhjrOhGerFI, it was created on default.scope-6.collection-12.