HAF instance communication stalled at some point
After starting an already filled HAF instance which was down some time (it had to catch up c.a. 200k blocks) it dropped indexes, performed massive sync, entered lived sync and started building up indexes again. This process again blocked processing for significant time (c.a. 2hrs). After that, node started to process blocks, but it seems it only consumed a part already filled into write-queue and even still had connected P2P peers didn't receive any block after that.
Maybe this problem is rather related to P2P communication layer.
Attached hived output.