Learning about behavior of complex system is most important reason in doing what-if Analysis. In system dynamics model we don't following up the optimization paradigm and we have a learning Paradigm and try to give knowledge acquisition from a complex system. Thus scenarios that would increase our conscious about system functional and structural behaviors are necessary for run in what-if analysis format.

What-if analysis is the better alternative to whoops-why analysis. It is about managing a system design process proactively, both with regards to optimizing performance and minimizing design errors. Systems dynamics modeling and simulation enables certain workflows across the complete design cycle to predict system performance and capture possible errors as early as possible using models, before the system is built. The system dynamics approach is suitable for any system with dynamic behavior governed by external stimuli or control systems. A general approach to what-if analysis would include to choose appropriate technology for the system modeling. One good choice is Modelica, which is an open-standard  modeling language that covers "all" relevant domains to design and compose engineered systems in a structured way. Using Modelica, the system can be described on component, subsystem, and system level, and a range of different analysis workflows can be applied for what-if analysis.

"what-if" just an additional analysis as many others. But it could be applied anywhere.

And It is the main benefit of "what-if" analysis.

I can understand your desire to have a bounded approach to asking such questions during system modelling and simulations.

in systems safety and security analysis, we use structured approaches to such questions often based on examining system model elements and thinking/asking what happens if/when the functions or the external transactions deviate from the expected norm. there's a standard set of such deviations for system safety/security/sustainability analysis that is applied consistently to all functions and modelled responses/transactions. can elaborate further if you are interested in the mechanics since others have explained the benefits sufficiently. In systems assurance, we need evidence of complete coverage of all system states to provide assurance/confidence about the safety/security or sustainability of the system in question

cheers