Comment 65 for bug 946663

Something doesn't make sense here. Aside from the fact that 'sleep' is generally a poor solution (how do you know how long to wait?), subshells on their own don't add any asynchrony: the only way that commands in a subshell could fail to exit before subsequent commands are run would be if the subshell in question were run in the background. That isn't the case here.

I think more investigation is needed. I'm confident that there's a better answer than inserting sleeps, but we won't know what until we know the true cause of this bug.