When I was much younger, I tended to think of writing along the lines of transcribing a movie that unreeled in my head. The book was a finished thing "out there somewhere", and my job was just to take dictation from it. It took a while to back away from this mental model, because as seductive and romantic as it was, it was also incomplete and sometimes downright harmful.
It wasn't until I started reading about theater, improv in particular, that I saw the better parallel was the creation of a sketch from improvisation. Many comedies on film aren't written word-for-word. They're stitched together from the best bits in a given scene across multiple takes, either in the editing room or by first running through multiple performances and then having the actors build a final composite from the best of the previous attempts. What looks spontaneous is almost certainly preplanned in some way, made to look like a (happy) accident.
The original idea I had about all this was one kind of "discovery", where you're Indiana Jones and your job is just to unseal the temple buried somewhere under your feet as long as you can find the right spot on the map. The misleading part of this mental model is that the work is intact and whole, waiting for you to find it in that form. It discourages you from experimenting with the work, from tinkering and trying and making attempts on things as you go.
I think one of the reasons people get attached to the discover-intact model is because it feels like a subtle refutation of the idea that you have to revise or rewrite. If you can just receive it all downloaded in one package, then that's that and you can move on to the next great download. But only the vanishing minority of truly good work comes in this form. We attach ourselves too freely to the romantic exception to the rule, instead of trying to see why the rule (revising, rewriting, experimenting) exists at all. It's like looking at lottery winners as the model to follow for how to invest wisely. (This is one of the reasons I don't think it's helpful to teach singular authors like Proust in writing workshops: as great as they are to read, they are bad models for how to learn to write, at least until you know why the rules exist and what benefits there are to their violation.)
The whole point of tinkering with something as you go is not to give yourself excuses to never finish it, though. George R.R. Martin once noted how one of the difficulties he encountered when working on the latter books of A Song Of Ice And Fire was how he'd come up with things he wanted to advance in the story, and then entertain far too many ways to put them in (dream sequence? flashback? footnote?). I suspect that in turn was a by-product of not having a narrow enough focus for the books to begin with, of having too few rules about what to put in and what to leave out. Narrow definitions, precise focus, make it easier to know what finishing looks like, and to what end the experimentation goes.
Drafts of a work are experiments. They just point the way, and help you narrow it further. The way, whatever it may be, is never as set as you think. It doesn't consist of an artifact that you have to dig out of its resting place and dust off. It's a path, one that you have to determine the best way of walking by simply walking it again and again.
Other Lives Of The Mind