We have a query (always the same query) which is from one of many (01-05) hosts. It occasionally shows up in our monitoring as having ASYNC_NETWORK_IO waits. My understanding was that this wait type is generally associated to the client not soaking up the data quick enough.
Given that's the explanation of that wait type, why does it cause our TempDB log to saturate? We get alerts through when the TempDB Log file fills up, I've left it until 15% free before. If we kill the query, TempDB Log instantly returns to 99% free space.
If all the query is doing is passing data from the output buffer to the client (albeit slowly because of the client application), what would cause the TempDB Log to fill?