Since SciPy 1.14 the method='COBYQA', which is supposedly a better method than 'COBYLA'.

Be sure to check whether this actually leads to faster runtimes, and doesn't negatively impact the accuracy too much.

Relevant methods

• lmo.diagnostic.rejection_point
• lmo.diagnostic.error_sensitivity, see also Vectorize diagnostic.error_sensitivity with scipy.optimize.elementwise.find_minimum on scipy>=1.15 #363
• lmo.diagnostic.shift_sensitivity

More info

• https://www.cobyqa.com/stable/
• https://docs.scipy.org/doc/scipy/reference/optimize.minimize-cobyqa.html
• https://theses.lib.polyu.edu.hk/handle/200/12294