What happened + What you expected to happen

We noticed that when our job was hanging with tasks producing no output (logs), tail_logs method was consuming 100% CPU.

While this is by design (awaiting new logs to tail them), this creates a wrong impression that this method is just in dead-lock busy-spinning.

This PR addresses that by

• Considerably simplifies it
• Prevents it busy-spinning

Versions / Dependencies

2.10

Reproduction script

NA

Issue Severity

Low: It annoys or frustrates me.