Work sits in your hands until you hand it over, and then it sits in a reviewer's until they answer. Four stages, two people, and no step that happens by itself.
Every layer of every file carries one. It starts as a draft and advances when somebody decides it should — never on a timer, never because a file was saved.
You are still working. Saves happen constantly and none of them tell anyone you are finished. Nobody is waiting on this file.
You said it was done. It is queued for a reviewer and locked while it waits — so what they approve is what they read.
A reviewer approved it. That is the end of the road for that layer, and it stops being editable for the person who made it.
A reviewer wants changes, and said why. It is open again, exactly as you left it, and you submit it a second time when it is right.
Solid arrows move work forward and are always somebody pressing a button. Dashed arrows bring it back — because you changed your mind, or because a reviewer wants another pass.
The only moment work changes hands is a submission, and the only moment it comes back is a reviewer's answer. Everything else is just working.
Open a file and work. It saves as you go, and Ctrl + S saves immediately. When the file is right, submit it — that is the handover, and the point where you stop being able to edit it.
If it comes back as rework, it is yours again. Fix and resubmit. There is no limit on how many rounds a file takes.
What is in review is what is waiting on you. Drafts are not your concern — that work is not finished, and nobody has claimed it is.
Accept it, or send it back for rework with a comment saying what is wrong. You can also reopen something already accepted; reviewers are not locked out of finished work the way labelers are.
These are enforced, not conventions. Asking for something on this list is refused with a reason rather than quietly ignored.
Submitting early is a normal mistake and it is not a trap.
Any file you submitted is yours to take back, right up until a reviewer answers. It returns to draft and you carry on where you left off. Nothing is lost and nobody is notified.
After a reviewer has accepted a file, the way back is to ask them to send it for rework. This is deliberate: approved work that anyone could reopen is not approved work.
Work on a draft for as long as you like, submit it when you mean it, and a reviewer either accepts it or sends it back with a reason — and nothing in between happens without somebody choosing it.