On the subject of mask planes, I note the Jim Bosch comment on Slack of 2018-02-05:
It now effectively means "this pixel was interpolated in one of the inputs and that's okay because we think the interpolation was good...
I think that's actually what `interpolated` was always supposed to mean, actually, but there were bugs in how we set INTRP and SAT in ISR that made it safest to flag on whether a pixel was interpolated rather than why it was interpolated. There should now always be an additional "why" flag set whenever INTRP is, and we should prefer to do filtering on those. For example, we're much better at interpolating CRs (can usually be trusted) than saturation (generally cannot).
This seems like something that should be included in the new documentation.