Subtle interactions in the embedded world - what bugs can teach us

> Each subchip has a full IMR; surely masking can be achieved by writing ~0 to all of them?

True. Maybe this comes under:

> or possibly we could make the implementation more clever in some way

The point is/was that the fact that masking at every level is needed, but that need isn't at all obvious.

Doing something more clever with the masking may indeed be part of the solution to meet that need.