I tend to agree with Kevin. If the operating case shows that there is lift off, I'll run a case with the support removed to check sustained stresses. For myself, as a compromise, if the load decreases, but doesn't go to zero, I take the sustained stresses with the support in place. I expect that, if the support has not actually been lifted off, then there will be some relaxation due to the secondary nature to bring load back on.

It is, perhaps, an arbitrary point, but it's the one I'm comfortable with. I think there is justification to say that there is some secondary nature involved, which is why I allow the support to "count" if there is still some load on in operation.

I've heard a suggestion for the cases where the load at a point goes from say 1000lbs in the SUS case to 100lbs in the OPE case. That is to make a model with a constant 100lbs force to replace the support point and check the sustained stresses that way. For me, that seems too extreme.