Seems it will be a significant quality jump on how tracks are handled.
While I see several BIG wish / requests that I consider important that are not fulfilled yet, I believe there should be more thought given to consolidation and internal optimization of BIAB. What pipeline is proposing makes perfect sense. At least on his model. Perhaps PG can make some kind of fast and dirty alpha / raw test version for those testers who are "not afraid" and willing to experiment?

If it can work as pipeline describing, it could speed everything up significantly and of course less file cluttering and possibly less tax on CPU.