Scheduler deadlocking

[hyunmoon]

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Connect all workers to lotus-miner (200+ of them)
2. Run lotus-miner sectors pledge
3. Find out all workers stay idle.
4. Run lotus-miner sealing sched-diag.
5. Process hangs indefinitely
6. I press ctrl+z.
7. There are 2 PIDs for lotus-miner.
8. lotus-miner does nothing no matter how long I wait.
9. lotus-miner stop doesn't work either. The processes have to be killed.

Expected behavior
NO JAMMING when calling sealing jobs list, sealing workers list, and sealing sched-diag. Please!
Jamming after running these commands seems to be the byproduct of sched deadlocking not the cause.

Screenshots

Version (run lotus version):
0.5.7+git.b8bbbf3e
(Compiled on latest mater and applied PR #3445.

Additional context
Sched has never been reliable from testnet phase 2 (when I started mining).
I've suffered from deadlocking issue for a long long time. Version 0.5.7 is bad enough to finally make me file a bug report.
The most reliable one I've tried was version 0.5.3. 0.5.7 is worst I've ever used.

[jennijuju]

comment from another miner that is related to this issue

[magik6k]

Can you get a goroutine dump when this happens (lotus-miner pprof goroutines > goroutines), and send it here?

[hyunmoon]

@magik6k
goroutines.log

I'll collect more dumps whenever I experience it from now on.