Following the discussion in https://indico.cern.ch/event/874111/, we decided to:
Still to discuss:
The full resolution model (without angular dependencies) would be reported by 6 variables, nearly doubling the size of the PrHit class. Alternatively, we propose in this MR to only propagate the pseudoSize information, and then to provide a helper class (or anything that would do it) which uses this information to deliver the most precise resolution model.
Presented and approved in https://indico.cern.ch/event/943209/