From Neo4j 5.18.0, in some of the cases the error messages as below would be logged in debug.log,

2024-05-14 14:55:36.754+0000 INFO [o.n.b.p.c.c.a.e.ErrorAccountant] Network abort rate has normalized
2024-05-14 14:55:36.881+0000 ERROR [o.n.b.p.c.c.a.e.ErrorAccountant] Increase in network aborts detected (more than 2 network related connection aborts over a period of 600000 ms) - This may indicate an issue with the network environment or an overload condition

This particualr set of messages have been introduced in Neo4j 5.18.0 as a replacement to the hard network errors we previously logged. Instead of logging every single error Neo4j will log this set of messages if there is more than the configured amount of network aborts.

Basically it means 

• Your network environment is unstable and connections are being dropped thus leading to these network aborts accumulating.
• You are killing driver instances without gracefully shutting them down (e.g. killing application containers or similar). If it's a one-off thing it can likely be safely ignored. If it persists, we should investigate those two causes and fix the underlying problem. This is not an issue with the server itself.

We made some changes to the way we report connectivity issues, and this is what you are seeing. 

To be specific, 

• If there is more than 2 connection aborts in a 10 minute window it will log this. It doesn’t mean there is something necessarily wrong but it should flag a concern, especially if you see many of these. 
• It is configurable, you can configure how sensitive the error accountant is and the window it reports over.

The related configuration in neo4j.conf as blew,
server.bolt.network_abort_warn_threshold
-The maximum amount of network related connection aborts permitted within a given window before emitting log messages (a value of zero reverts to legacy warning behavior)
server.bolt.network_abort_warn_window_duration
-The duration of the window in which network related connection aborts are sampled
server.bolt.network_abort_clear_window_duration
-The duration for which network related connection aborts need to remain at a reasonable level before the error is cleared.

There's a variety of errors such as "Failed to write response", "Connection Closed", "Connection Timed Out", etc. that would've previously been logged instead of this message”

Reference:https://neo4j.com/docs/operations-manual/current/configuration/configuration-settings/#config_server.bolt.network_abort_clear_window_duration