- Optimized message queue handling, now queue cleaning speed doesn't depend on total queue size
- Cleaning delivered messages using lt augmentation instead of random search / consequtive walk
- Keeping root cell of queue message in memory until outdated (caching)
- Changes to block collation/validation limits
- Stop accepting new external message if message queue is overloaded
- Introducing conditions for shard split/merge based on queue size
Read more on that update.