Steps to Reproduce:
1. Install 25 nodes pool and fill ledger with 200k+ NYM txns.
2. Run 800 writing (4 machines with 200 threads from each) and 2000 reading clients (10 machines with 200 threads from each) that write/read 1 NYM txns each second.
Nodes wrote 30..40k txns until the whole pool stopped writing and reading with PoolLedgerTimeout error. The most of nodes have 252k txns and the most lagged node has 222k txns written.