500 rate necessary?

[finnhodgkin]

The code looks good to me, except possibly 500s being included. I thought the app issues were caused by timeouts from dyno freezes, not normal server 500s. I would have thought that if a request gets to the point of a 500 it means the dyno is healthy enough to get to the error and probably doesn't need a restart.

Slightly worrying because dyno restarts could be triggered from completely intentional error handling. Also I reckon after this has been in production for a while we're quite likely to forget it exists so debugging could become a bit of a nightmare if normal errors are also included.

[roryc89]

Quite possible this could be an issue. I don't think we should intentionally be responding with 500s though as I don't think we should intentionally error. It should be a 400s client request issue if the request isn't valid.

So far this doesn't seem to be an issue and I think it's possible that at times a dyno restart could stop 500s.

Also we can see in the main app metrics when the dynos have been intentionally restarted.