Why is my application getting bursts of H19s when it didn't before?

Issue

You're seeing bursts of H19s spread out between your dynos that come and go periodically over the course of a day. Previously, you never saw errors like this coming from your application – even under higher traffic loads.

Resolution

Bursts of H19s like this can happen when Heroku routing nodes are undergoing maintenance. That typically either takes the form of restarts or node replacements. This is done as part of our normal operational procedures, which is why you won't see posts on our status site indicating that this is happening.

If you're seeing this repeatedly, then it may be due to rolling restarts being performed on our routing fleet to address an issue (e.g., pushing out security fixes for a recently published CVE). That process typically takes around one week due to the slow rate it's performed to minimize impact. It can also happen repeatedly if we're dealing with bad routing nodes that require replacement. Usually, bad nodes are isolated, but there are occasions where we can have several over the course of a week.

If you see bursts of H19s for a longer period than a week without it returning to normal levels or frequency, then please contact support!