How can data capture be combined with physical simulation, or what does the workflow look like with these two methods?

Most likely these are separate steps - you would perform data capture first and then simulate by inserting the captured rigidbodies into the scene. A real time visualization of this might be cool/practical for people working on set though?

I think this motion capture use case at Boeing is really neat: https://www.boeing.com/company/key-orgs/boeing-technology-services/additional-services.page

Using motion capture to estimate joint stress and fatigue when performing a task.