While I agree with the statements that MLR models, “do not account for future mutations and can only project forwards from circulating viral diversity. This intrinsically limits the effective forecasting horizon achievable by these models.” It seems to me that the discussion should center in the purpose of the models to begin with. So, for a more balanced discussion, I would suggest the authors to add perhaps a paragraph in the discussion on the role of adding mechanism or background information (vaccine, previous waves, to understand immunity profiles of populations) onto these models, if they are to be useful beyond the nowcast, as it will no doubt affect internal dynamics and competition between variants that in turn affects sustained growth and displacement of other less fit variants.
TODO:
Revisit this part of the Discussion. General framework still useful to measure performance even if there is mechanism on top of MLR. Mechanism could project further. Need to include both mutation and mechanism to project farther than 30 days.
Longer term prediction will have changing fitness, role of mechanisms, and then something about extending.
TODO: