IMO this kind of panic should not be recoverable. I guess I'm arguing for
runtime.throw vs runtime.panic

On Sat, 9 Jan 2016, 04:23 Roger Peppe notifications@github.com wrote:

Not necessarily bad behaviour but unexpected:

http://play.golang.org/p/OrPRAXnkdA

I would expect the program to complete here (because
we've recovered from the Unlock panic so the mutex
should not be locked) but it actually
deadlocks before the failed unlock locked the
mutex.

—
Reply to this email directly or view it on GitHub
#13879.

This issue has nothing related to channels, it is just that you can't lock a bad mutex.
Here is a simpler example without channels: https://play.golang.org/p/Xt76u_Tm7w

CL https://golang.org/cl/31359 mentions this issue.