Old data is still present in dataset after upgrading from 6.6.2 to 7.0.0

Pre Upgrade Steps -
1. Load bucket into KV.
2. Load docs into KV
3. Create multiple dataverses.
4. Create multiple dataset on the bucket.
5. Create multiple indexes on datasets.

Upgrade the nodes using online upgrade with swap rebalance.

Post upgrade steps -
1. Rebalance again to activate CBAS service.
2. Validate metadata entry for all the dataverses, datasets and indexes created before upgrade.
3. validate doc count in the datasets created before upgrade.
4. Validates indexes that were created before upgrade.
5. Load some more doc in default collection of default scope of the bucket created before upgrade.
6. Check the newly added docs are ingested into the datasets created before upgrade.
7. Delete (and not flush) all the data from the buckets created before upgrade.
8. Create more buckets, scopes and collections on KV side.
9. Load docs into all the buckets.
10. Create some more dataverses, datasets, indexes and synonyms.
11. Wait for ingestion to complete.

For datasets created before upgrade, the number of docs does not match number of docs in associated KV collections. The dataset still has old data.