would be similar to what HILO did back in days of yore ...

There could be a schedule file, taken as stimuli to the simulator. It would specify when events external to the MCU occur, and would then allow the simulator to respond to those inputs. Outputs and their interaction with inputs can be simulated via logic equations, providing feedback paths if any exist. Multiple stimulus sets can be employed.

These external files can take the form of external memory content, external schedulable event sequences, including recurring ones, such as a 60 Hz interrupt, and the like. They can also take the form of external devices characterized in rigorous behavioral sequences. I think HILO did that with an interpreter that generated its outputs on the fly as the simulation ran.

The net product, in the case of HILO, was that one could specify a set of stimuli, and subject a physical device to those stimuli, thereby generating a behavioral sequence. The output from that exercise could then be used to characterize, for purposes of simulation, the behavior of that device in subsequent simulations. This was not an exact modeling but one limited to the stimuli to be used in those subsequent simulations.

Needless to say, it's not a simple task, as limiting a simulator to one specific device might be, but it is finite and, therefore, achievable.

RE