Race condition in index-batch processing

[jleandroperez]

Scenario:

1. An odd number of objects is sync'ed (+50), after login.
2. The index is expected to be processed in batch (50 changes per batch).

Note that those objects that cannot fill an entire batch (less than 50 changes) should be processed as long as the last pending change arrives.

Found a race condition in the batch-handling logic, that will cause groups of less than 50 changes not to be processed. Let's get it fixed.