I'm afraid there's a race there: we're starting 30 goroutines but immediately call for docker events, I don't think we have any guarantee to reliability get the 64 expected events in the end.

I'd add a sync.WaitGroup: @LK4D4 what do you think?

Also can't use dockerCmd in a goroutine like that since it calls c.Fatal on error.

Better to use directly runCommandWithOutput ?

Yes, and any error would need to be dealt with from the main goroutine.

Better synchronize everything with a channel to handle eventual errors from dockerCmd runCommandWithOutput and I'd use another to handle docker events streaming checks

Thank you for your hints :)

should be really buferized to 30, to let goroutines just die.

nit, I'd call it errChan since errors could be imported (again, nit, don't necessary do it and maintainers have last word of course, wait for them and in case make this change)

Yeah, errors is package from stdlib, so, can be confusing.

Yeah, thank you, I'm fixing this

Ok, all done!