Handling large messages, do we drop them? Log them to disk? Callback in confirm?

[haf]

For example, we have Rutta ingesting messages over HTTP and once in a while someone sends a large message >134 KiB, which causes the Stackdriver target to crash and restart, until all too-large messages have been crash-rejected (since we don't have a "will" in that target).

[lust4life]

what is the use case? seems disk or queue are something like target. is it for when target can not catch up with log message producer ?

[haf]

Updated description