@centril was asking whether we ought to backport #58649 (cc @pnkfelix)
(it ICEs on beta)
I guess that ICE is a regression?
Seems like a pretty safe backport
delay_span_bug is basically zero risk
@nikomatsakis gave you r? on the PR
@nikomatsakis I can p=100 it, or do you want to wait for travis to go green first?
I see no downside to backport