Currently Jenkins can recover from OutOfMemory errors in particular cases, but there are known issues like channel hanging.

They issue may stay unnoticed till it disappears from logs. In such case Jenkins admins may have no clue if it was a cause of connection hanging.

Proposals:

• Record last OOM events in support bundles (if they happen after the restart)
• Add an administrative monitor with a warning (Jenkins restart is recommended in general)

It may likely require patches in the Jenkins core