As described in 

https://bugs.chromium.org/p/chromium/issues/detail?id=587247#c46

I think there may be (or at least was) a potential race in GN that might cause errors and/or nondeterministic results. 

We should see if we can reproduce this or if it happens again. I could theoretically reproduce it by syncing to before r397471 and applying the patch in 

https://codereview.chromium.org/1997493002

(as per msg #11 on that CL).