Pool has stopped to write txns

Build Info:
indy-node 1.5.526

Steps to Reproduce:
0. Install pool of 25 nodes.
1. Set client stack restart parameters:

MAX_CONNECTED_CLIENTS_NUM = 10
MIN_STACK_RESTART_TIMEOUT = 600  # seconds
CLIENT_STACK_RESTART_ENABLED = True

2. Run several load tests changing CLIENT_STACK_RESTART_ENABLED between True and False:

python3 perf.py -c 10 -n 10 -t 1 -g pool_transactions_genesis
python3 perf.py -c 20 -n 10 -t 1 -g pool_transactions_genesis
python3 perf.py -c 30 -n 10 -t 1 -g pool_transactions_genesis
python3 perf.py -c 10 -n 10 -t 2 -g pool_transactions_genesis
python3 perf.py -c 15 -n 10 -t 2 -g pool_transactions_genesis

3. Restart the whole pool several times.

Actual Results:
1. Pool has stopped to write txns with PoolLedgerTimeout error.
2. Pool is unable to perform catchup.
3. 17 of 25 nodes have 326349 txns written.
4. 8 of 25 nodes have 326359 txns written.
5. Validator-info shows all nodes as reachable.