Upsert and Indexing discrepancy

I was doing a larger run and I did notice that after I stopped upserting, the number of vectors in index kept on increasing. Eg. I completed upserting around 7 AM (total vector count ~40M). After 3 hours, the vector count continued to go up (from 40M → 45M). This was using gRPC index.

How long is the delta between upsert request completed vs all vectors are indexed and available for query?

The time it takes for the index to catch up really depends on the size of the queue and the number of pods so there is no definite answer I can give you. We are hoping to expose a metric on this for advanced users in the future.