Well I’m not so sure about the answer to that. Another thing that can be done, instead, would be to plot a range of solutions near the answer, and test results of known similar conditions.
I mostly did work like this. When something goes awry you’ll be able to tell most of the time. Complex real and imag values become underdetermined you’ll see steppy swaps. Really though you can either measure or find the floating point error for most calculations. Julia is actually amazing for this… I need to write a paper about it - but have negative time to do so. I have tons of results and code for it though…