Separate on-the-fly variation PS weights for different splittings
More and more experimental analyses are using template/profile likelihood fit techniques to constrain the MC modelling (basically constraining a signal region from a control region, but more dynamically/differentially).
To avoid being too aggressive, it would be important to have separate on-the-fly scale variation weights for different splitting kernels, which can then be decorrelated in the fit. For example, it would be good to have separate variations of g->QQ, g->qq, g->gg+q->qg. Potentially for the same reason it would be nice to have separate weights for IS and FS splittings.
I am opening this ticket to collect opinions from users on what division should be done, at the same time avoiding a proliferation of weights, and to explore whether we can already include this in a short timescale in a Sherpa 2.2.8 release.