I think you actually just described two of the things I was trying to say:
So one of the points I made that I think you misunderstood has to do with the above.
You write automation and then - after - you want to experiment; so you turn “write” off. So now your automation will recall whatever was written before (and if you’re before where you wrote automation it’ll pick up the stuff at the beginning of the timeline). You press play, and you “try something” with your automation. To then write it, you “re engage write to have it written”.
My point earlier was this: Suppose you do exactly what you described, but perhaps something happens - the phone rings, the pot of water is boiling on the stove, whatever - so you choose to not yet punch in by “re engaging write”… IF you then stop or let the playback either reach automation written later or cycle back, THEN the settings you “tried” will be lost, because it will again read automation. IF it did NOT read automation on stop/return (cycle), you would still have those last settings you “tried” in VT. This would have been a different way of working, and one that is arguably more intuitive when just looking at the automation written on lanes. Not necessarily better, but different.
That’s what I was saying earlier. Your intuition was exactly the same as mine, and it appears other users expect the same thing. It’s not “complicated”, it’s just not intuitive. So from that standpoint it’s confusing.