Metrics endpoint has no timeout — response time grows unbounded under load

[kriszyp]

Problem

Under sustained high load (100 concurrent users, no think time), the GET /prometheus_exporter/metrics response time grows indefinitely — reaching 483 seconds within 8 minutes and still increasing. No timeout is enforced on this endpoint.

Fix

Add a request timeout (e.g., 60 seconds) to the metrics scrape handler so stuck requests are terminated and resources are released.

Status: In Progress in Jira.

🤖 Filed by Claude on behalf of Kris.