At present the model of the background that has been subtracted from an exposure is a separate entity from that exposure. It can be difficult to keep the two objects synchronized as backgrounds are estimated and subtracted, and persistence is a nuisance because it requires two separately managed dataset types.
I propose that the model of the background that has been subtracted be kept with the Exposure and persisted with the exposure.
This RFC is primarily intended to get permission to make this change. I am not sure we can implement this proposal yet. Everything else in an Exposure is a C++ object, and background models are Python, but my understanding is that we will be able to solve this eventually.