When the local cluster uses 2PC but the inter-site replication is configured to use 1PC, during the local prepare, the number communication failures with remote site is reset to 0 - that should not happen.