since this works based on extrapolation it should pull back all controllpoints in a row, so the amount of cps and the knotvector shouldn‘t be a problem. Well I never use this with multispan surfaces, so I dont know if extrapolation works just for each local segment then. Its worth to test. That wouldn’t be a problem then…
Furthermore the limitations are obvious: Your cutting operations needs to allow 4 edges and your cutting shape should be clean, so that the amount of controlpoints can approximate the intersection. Deviation is acceptable. What it is more about is to tangent align edges in a surface layout, in order to improve the conditions to match a surface. So real trim/surface splitt(vsr terminology), is heavliy used in automotive modelling. I use this all the time. Especially for corner fillets this is a very useful tool.
A thread from last week:
in this thread I used this functionality, to create the blend. Unfortunate I wasn‘t properly splitting, but still, the conditions were better than without this splitting operation. I also used rational fillets