A friend shared this link to a comic-style presentation of the difficulties inherent in modeling a pandemic. In nutshell, lots of potential variables (not all of which are necessarily relevant in all contexts), and not enough data to populate those variables. Ergo the predictions arising from those models may be wildly off.
For many patent practitioners, there's a familiarity there, that comes from dealing with a prior art document that throws in everything but the kitchen sink. It's not unusual for some patents or published applications to provide long lists of the potential identities of various components in a pharmaceutical or other formulation, or in a process for making a chemical compound or a formulation, or to provide long lists of potential substituents at various positions of a molecule. Or maybe the patent recites broad ranges for the amounts of various components in a formulation or a process.
When all those variables are taken into account, there may be millions or hundreds of millions of possible combinations. Even the best-funded applicants can only provide a few hundred examples to illustrate those millions of possible combinations, and more often than not, the application will provide only a handful of examples. And yet for a patent examiner - speaking, of course, for the hypothetical person of average skill in the art - when looking at that earlier publication as prior art, it was always obvious to pull out the much narrower group of combinations or the specific combination that's claimed in your client's own pending application.
Sure. Just like it's easy to model a viral pandemic.