Usable results are thrown away if recoverable errors are encountered

[biotinker]

See the code here in Optimize:

	var cResult C.nlopt_result = C.nlopt_optimize(n.cOpt, (*C.double)(unsafe.Pointer(&cX[0])), &cOptF)
	status, err := normResult(cResult)
	n.setStatus(status)
	if err != nil {
		return nil, math.NaN(), err
	}
	return toGoArray(cX), float64(cOptF), nil

If there is any sort of error encountered, then nil, NaN is returned rather than what nlopt returned.

However, nlopt will frequently return usable values in addition to an error; for example, if the optimizer runs out of floating-point precision prior to hitting an exit condition, it will error with "ROUNDOFF_LIMITED" and return the last value, which usually is a good, usable value. Unfortunately, this value will be thrown out.

I understand that the "Go way" is to return nil and an error, or a usable value and a nil error.

However the use of that paradigm here destroys useful information. Is there a way to resolve this?