Issue Summary

Confluence Data Center 10.2 can become unresponsive when a read-through cache loader waits for an unusually long database operation. 
 
Tomcat’s stuck-thread detector may initially report a single long-running request:

WARNING [Catalina-utility-2] org.apache.catalina.valves.StuckThreadDetectionValve.notifyStuckThreadDetected Thread [http-nio-8080-exec-617 url: <redacted> has been active for [64,542] milliseconds (since [7/28/26, 2:27 PM]) to serve the same request for <redacted> and may be stuck (configured threshold for this StuckThreadDetectionValve is [60] seconds). There is/are [1] thread(s) in total that are monitored by this Valve and may be stuck. 

This can then lead to other threads becoming stuck and the number of long-running requests increases:

...and may be stuck (configured threshold for this StuckThreadDetectionValve is [60] seconds). There is/are [13] thread(s) in total that are monitored by this Valve and may be stuck. 

Eventually, this can lead to exhaustion of the HTTP thread pool and the node becoming unhealthy. The node recovers when the stalled cache loader completes or after the affected node is restarted. 
 

Steps to Reproduce

To be confirmed

Expected Results

Node remains healthy

Actual Results

Node becomes unhealthy with HTTP thread pool exhausted

Workaround

Currently there is no known workaround for this behavior. A workaround will be added here when available