It happens too often. I come upon a server that the customer has called in as being unstable while running the Web server. I come in, and
the domlog is huge. I mean lots of gigabytes of data.
The domlog is unused. That's mostly because it can't be opened, so much data's being added, the view update can't even happen.
The domlog is impossible to analyze. Again, it'd require a whole other server to get the thing's views open, but scanning it for even a bit of data proves practically impossible.
One tradeoff suggestion: store the messages more like the log.nsf. Put multiple messages in a single doc. I'm serious. I 'm noncommittal about this format because it's not ideal, but at least it shields the server from having to update views with 100k-Ms of docs. Plus, Development probably already has the code to write to "another log.nsf". That would be far preferred over my situation, where the log takes hours just to open it.
I could think of other possibilities, too. Highly specialized view indexes & collections, that may eventually take their places as new view types for high-performance applications. Hey, I can dream.