Well, yes, a bulge in the middle is actually 4th order: 2nd order in X and 2nd order in Y. So I modelled that and tested the Solver on it. Failure.
It was an interesting exercise all the same. The Solver was designed for business use, as far as I can see, which means it is focused mainly on linear models. It works quite well (very well) for those. I know that a full conjugate gradients package can handle 4th order easily enough because I have written such a package in the distant past, and used it very successfully.
Unfortunately the Mach Formulas feature will not handle the problem anyhow, as it is a very half-hearted implementation. It can handle G0 and G1 to a limited degree, but not G2/G3. Ah well, the real solution is to build a proper vacuum table - which is what my wife says I should have done in the first place.
cheers
Roger