Part of nodes continued ordering txns after `incorrect state trie` under load

Environment:
indy-node 1.4.463
libindy 1.4.0~565

Steps to Reproduce:
1. Setup the pool of 25 nodes with following parameters in config:

DELTA = 0.1
LAMBDA = 240
OMEGA = 20
Max3PCBatchSize = 10000
Max3PCBatchWait = 1 

2. Run the load test:

python3.5 perf_processes.py -n 1 -t 1 -c 10 -r 100 -g perf25_transactions_genesis

Actual Results:
Pool have wrote ~975K txs during several days. First View Change was happened on ~950K txs due to primary node failure because of out of memory. After that one more view change was happened and `incorrect state trie` appeared on part of nodes. But nodes with `incorrect state trie` were continued txns ordering.