Currently if I stop the cascade in the middle of thinking, reverting to last request may lead to loss of original scripts (which weren't be edited in this halfway thinking). The irreversible nature of reverts makes it a heavy loss of progress.
o
orioto
Normally, what i do is stage all changes files after each message if the result is good so I protect it from loss if your case is happened, and continue then commit after the task is done