Here's what I have done successfully in a situation like that: I like putting epoxy on almost immediately after blasting to bright metal. I do all my work outside in the summer and live in a fairly humid area. Even fingerprints on a bright-blasted steel surface can start the rusting process. Getting epoxy on as soon as possible protects the metal for additional work later on. I've also used Picklex 20 for that purpose but prefer epoxy if I can accomplish it at the time. If there are some areas that need plastic filler later on (hammer/dolly work should have already been completed), I'd rough the area with a course grit (say 80). That will go to bare metal in the heavy scratches. I then would apply a fresh coat of epoxy to the area and apply the filler within the epoxy window. After sanding the filler as needed, there may still be spots where sanding went to bare metal. Scuff the area at that point (Scotch-Brite is good for that) and apply a final coat of epoxy. You then could go directly to the topcoat within the epoxy window if the surface is smooth enough at that point or apply a few coats of surfacer. You then could delay the topcoat for a time of your choosing without worrying about any further time windows. You'll need to final sand the surfacer anyway and that will provide the "tooth" for the topcoat.
I have never had a problem applying filler over epoxy that way. I'm not questioning that filler to bare metal with a heavy scratch is "better" - and I've seen some test data that so indicates. However, I remain unconvinced that it's necessary in the vast majority of cases. So I prefer epoxy first, then filler.
I know that B- prefers to add a fresh coat of epoxy between the final-sanded surfacer and then topcoat within the window. Not as any kind of sealer - no sealer is required. The coat of epoxy at that point provides for a chemical bond with the topcoat and is "better" than just the mechanical bond using only the scratches from the surfacer sanding, usually with a final 320 grit or so. I've never had a problem skipping that step and feel that while it is "better" it's likely not necessary considering the more applications you do, the more chances for something messing things up.