Please take into consideration what context is passed here when this code is used. On shutdown the context will likely be cancelled but we still want to try to send all the logs.

Do we actually need to decrement this?

I mean, it's probably fine, but in theory I supposed we could end up queueingmaxBytesQueued even after the database is near-full. Let's say we sentmaxBytesQueued-1, then the connection is interrupted, and we queue up anothermaxBytesQueued only to be discarded after connection is re-established.

I didn't want to reimplement the server side limit in the agent. So, if the server changes the limit the agents would still work with the new limit.

I get that we want to havesome agent side limit since the logs are stored in memory, and so the current server limit feels appropriate. But, as long as the agent is keeping pace relatively well sending logs to the server, I don't want to stop it sending until the server says we've reached the limit.

Since we don't give an error, perhapscontinue is better here so we don't discard the entire batch?