I recently landed two CLs:

1) https://codereview.chromium.org/2649973003/
[Add files A/B]
2) https://codereview.chromium.org/2650363002/
[Update A/B]

(2) is dependent on (1). (2) caused a compile failure on an ASAN bot, but the sheriff incorrectly reverted (!). To everyone's surprise, the revert landed:

3) https://codereview.chromium.org/2658083002/

However, if you look at what rietveld claims is contained in this CL [both by diff, or just file by file], it's different from what git claims was landed:

See hash e9942e23fec9e2bca8ebd70f79feb2120943b54b.

I think what's happened here is that rietveld uploads a diff, which may or may not represent the actual patch that we try to land on trunk, which presumably just deletes A and B, without looking to see if the diff matches. This seems dangerous, since (2) has effectively been reverted, but it's very hard to tell that this is the case from looking at rietveld.