Explanation of reconnection/renegotiation

[philipflesher]

We get a lot of errors in our logs as nodes drop off and are attempting re-establishment. Undetermined at this point whether this is a Thread issue or a Matter issue. I suspect most likely Thread, as we have a large deployment.

One thing that would help is a detailed understanding of how the Matter server does its reconnect/renegotiation of nodes, including its liveness checks, how often it decides to ping nodes, when it decides to retry, how many times it will retry, and what happens after it reaches maximum retries (give up for a while? give up forever, until device proactively reaches out? something else?). The behavior we see is that a node will be offline for many hours without coming back, and then for whatever reason magically reappears later.

I can provide logs to help showcase the specific situation, if that's helpful. But really I'm just looking for any doc or an explanation so I can educate myself and further diagnose things on my end.