Most probably it is an intermittent failure of Prometheus metrics endpoint with HTTP/2 and newer Jetty introduced with 7.13.2.

A critical issue has been identified where the Prometheus metrics exporter endpoint on the broker intermittently stops responding.

Despite this failure, the core broker functionality remains operational, handling message
traffic and other tasks without apparent disruption.

Furthermore, the metrics port remains open and accessible at the network level. So no readiness failure.

However, the internal mechanism responsible for serving the metrics data (the endpoint's back-end) ceases to function, resulting in an empty response when Prometheus attempts to scrape the data.

Crucially, this behavior is immediately resolved by a complete restart of the broker.