When the working hypothesis is wrong, what next?

A fully formed hypothesis about a system lends itself to quantitation and then can be expressed as a model, in precise terms, e.g. PV = RT, the pressure times the volume of the gas equals the gas constant times the temperature). In order to merit the effort to disprove the working hypothesis it must be precise. Models used for the analysis of physiological data should be detailed enough to describe the results of at least a few different types of experiment. But as the experimental data become more profuse and more accurate the model fails; some of the data are systematically misfitted; the model is WRONG. What next? Platt's (JR Platt Strong Inference. Science 146:347, 1964) admonition is: develop alternative hypotheses, design and do the experiments that distinguish between them, beyond the noise in the data. At least one hypothesis will be condemned to the dust bin. And then one is in the iterative loop of experiment, disprove working hypothesis, find alternatives, and repeat the process. No model can be proven correct; but a good working hypothesis is worth a lot in providing understanding and prediction. Finding the alternative hypothesis is the central mission. The model failure is the key to advancement of science, therefore attack your model at it weakest point and prove it wrong. Examples abound in astronomy, physics, chemistry and biology.