v0.71.3-rc.2 : Unexpected protocol errors in Longevity test "NetworkUtils: Connection broken"

Unexpected protocol errors in Longevity test:

[https://perf.analytics.eng.hashgraph.io/ephemeral/test_runs/v0.71.3-rc.2_AdHocSD8_537/report[…]ty/podlog_solo-sdlt-n8/network-node2_logs/swirlds.log](https://perf.analytics.eng.hashgraph.io/ephemeral/test_runs/v0.71.3-rc.2_AdHocSD8_537/report/longevity/podlog_solo-sdlt-n8/network-node2_logs/swirlds.log)
```
2026-03-09 20:51:47.818 70111 WARN SOCKET_EXCEPTIONS <> NetworkUtils: Connection broken: 1 -> 2
org.hiero.consensus.concurrent.pool.ParallelExecutionException: Time thrown: 2026-03-09T20:51:47.616053850Z
...
Caused by: java.net.SocketException: Connection reset by peer
```

[https://perf.analytics.eng.hashgraph.io/ephemeral/test_runs/v0.71.3-rc.2_AdHocSD8_537/report[…]ty/podlog_solo-sdlt-n8/network-node3_logs/swirlds.log](https://perf.analytics.eng.hashgraph.io/ephemeral/test_runs/v0.71.3-rc.2_AdHocSD8_537/report/longevity/podlog_solo-sdlt-n8/network-node3_logs/swirlds.log)
```
2026-03-09 20:51:45.962 70105 WARN SOCKET_EXCEPTIONS <> NetworkUtils: Connection broken: 2 <- 1
org.hiero.consensus.concurrent.pool.ParallelExecutionException: Time thrown: 2026-03-09T20:51:45.955108053Z
...
Caused by: java.net.SocketTimeoutException: Read timed out
```
It appears, though not necessarily so, the two nodes are waiting for each other until timeout breaks the deadlock.

Start with the referenced swirlds.log entries from the Longevity test and trace the NetworkUtils warnings in the SyncProtocolWith2 and SyncProtocolWith1 paths. Compare the node 1 and node 2 connection-reset and read-timeout events to determine whether both sides are waiting on each other. Done means the cause is identified and the unexpected protocol errors no longer occur in the relevant longevity run.

Written by the indexing model from the issue text.