Incremental indexing progress is too abrupt

Steps:
1. Disable automatic index updates.
2. Create dataset: 1 bucket, 10M x 2KB items
3. Build initial index (4 ddocs, 10 views per ddoc)
4. Double dataset
5. Trigger index update

20 minutes of 0% progress for given case, from 78% to completion. See attached log for details.

With smaller datasets it goes straight from 0 to completion.

This is pretty much know issue but didn't find any related ticket for that.