Yes, other kind of surfaces families would be wellcome.
Landslides do not have excactly circular sections, but more like a logarithmic spiral.
But all current software I know about uses still circular surfaces. One thing I would like to see is generating this kind of surfaces and comaparing results with the traditional aproach.
Of cource you need a precise method like spencer or morgensten-price (spencer is just morgenstent-price with a constant function) in order to simultaniously satisfy moment and force equilibrium, as the movement won't be a translation (as planar slidind where jambu may be satisfactory) or rotation around the center of the circle (where bishop will be adequate).
But you can provide results any way from these aproximate methods even if the results won't be meaningfull.
Other kind of surfaces like poligonal lines (along contacts for example) genarating arcs or straight lines in the extremes to find the surface would be very usefull in other situations (like in slide program or geoslope).
These programs provide results for all the methods when you select non circular surfaces, even if the results won't be acurate and theoretically not very meaningfull.