Thanks. 
My situation is this: I’m writing a course for second-year biology students on using dynamical modelling to model cellular, physiological and ecological system, and am intending to use DynamicalSystems as the basis for the course. Students will need to debug their dynamical models, and so will need to investigate how features of their model change dynamically over time.
For example, when simulating glycolysis, they might want to explore how oscillations arise by calling trajectory() and have it return a history of how two different flows du of material between two compounds ADP and F6P vary over time. Or they might want to define their own variable energy, have it updated on each step, and study its history after a run to see how it varies.
I realise that I could have them step!() through the simulation to update variables, but I’m assuming it’s more efficient to use trajectory(), and in any case, I want the students to concentrate on the biological content, rather than on implementation details. I also gained the impression from your reply to another post of mine, that it would be an abuse of their intent to store changing values in a parameter.