A number of users have observed that, when surface-attaching parts such as solar panels and RCS thrusters onto the smaller-diameter fuselage of the Service Module's propulsion segment, said parts sometimes align with the wider avionics ring segment instead, leading to the illusion of parts "floating" a distance from the Service Module.

This was first observed in the transition from 0.25->0.90, and I suspect it may have something to do with the Service Module colliders (or possibly how KSP handles collider physics).

I've updated the Service Module colliders, but results have been mixed. It may be that some parts have bounding boxes that are too big to be used with SDHI.

Not much I can do on my end I'm afraid.

Users have reported that the issue is absent (or at least less severe) in the V1.9 version of the Service Module.

After reverse-engineering the V1.9 model, I took out the convex colliders in the upper ring of the Service Module, thus (mostly) reverting the surface attachment to the behaviour seen in V1.9, but keeping the umbilical.

User feedback seems to be promising, but I'm waiting for further data.

Further feedback from users:

• There are a few spots wtih abnormal surface attachment orientation; however, these reports concluded that they were edge cases that can be safely disregarded
• Users felts that the trade-off of having non-convex colliders capable of proper interaction with terrain was acceptable, considering that most of the time, the Service Module will survive ascent and later burn up during re-entry as designed.

I think it's safe to say that this has (finally) been fixed to everyone's satisfaction.