I checked this, and here are my results:
My tests were rendered in Max, saved to 32-bit EXR, then composed in Fusion.
Beauty and alpha, then checked with black background, cyan background, and the actual HDRI background.
In all cases, the result with adaptivity enabled is better, or the difference is negligible.
Also, without adaptivity, other areas are getting more noisy - such as reflections in glossy objects.
So to me it seems that everything is working fine - adaptivity helps focus samples where they are most needed, so yes, the dark part of the shadowcatcher will become noisy, but:
- it doesn't matter, since we won't see those areas anyway
- you never place shadowcatcher objects against a fully black or fully white background (however in my examples it didn't matter anyway)
I am waiting for some more specific information how this could be reproduced.
I have some images, but I am not sure if I can share them (Francesco - can you please confirm?).
Looking at the bump spinner issue now...
Update: the bump spinner seems to work fine, too. Waiting for further instructions.