First Pre-Prepare message has incorrect state trie root right after view_change (on master replica)

Steps to reproduce:
1) Restart pool via CLI
2) send transactions via perf_processes.py

python3 perf_processes.py -g pool_transactions_genesis -b 10 -r 2 -m t -c 2 -n 1 -k nym

Actulal results:
6 of 25 nodes stopped writing

 Logs will be available on LogProcessor