Is it possible to train on a set of series ?

[NonAbelianCapu]

Hi! First of all, I wanted to say that the library has been really useful for my PhD, so thanks to the developers for creating such a cool open source project.
Regarding my question: in my PhD I work with non-ergodic systems where a single realization is not informative, so a meaningful estimate requires pooling information across many independent trajectories sampled from the same stochastic process. I'm aware of the Monte Carlo pattern in the docs (looping .fit() over realizations and aggregating theta afterwards), but what I'd ideally want is to pass a list of trajectories directly to the estimator so that the regressor matrix is built jointly across realizations before solving — which would also affect the ERR ranking in FROLS, not only the parameter estimates.
From my reading of the API this isn't currently supported — am I correct? If so, would the maintainers be open to a contribution along these lines?
Thanks so much!