23.4.1 node running in docker is showing in other peers tables with ?discport non-zero

[matthew1001]

Description

A 23.4.1 besu node running as a non-validator in a QBFT network is showing in other peer tables with an enode URL that includes a ?discport=X (where X is a non-zero port number).

The node is running in docker with either --nat-method=NONE or --nat-method=DOCKER (both have been tried, and discport still appears to be set to a non-zero value)

To some of the validators, the enode URL does not contain a discport query param and those validators can connect to the node correctly.

To other validators the enode URL does contain a discport as mention above, and these validators cannot communicate with it. Trace shows failed PING packets to the node on that port.

Expected behaviour

The node isn't expected to have separate UDP discovery and TCP listener ports. The --p2p-port CLI param has been set, and is used for the TCP port. It should also be used for the UDP port. Docker is run with port porwarding configured for that port.

As far as I can tell, a node would only advertise a separate discovery port if the node was running in Kubernetes, and a LoadBalancer was defined with a port named discovery. So it's unclear why the enode has a non-zero discport, and also unclear why some validators show the URL with a discport and some don't.