Training may take very long, could make sense to some, save the training already done.
On training same strategy again, ask if startover or continue where left off.

From the other hand if Zorro S, then one could just open another Zorro instance. No?

Just was thinking on costly calculations, e.g. training neurons and things.
Save the efforts achieved is in general always a good idea in IT.

Training neurons is not done by Zorro, but by external libraries. I do not know a library where this can be interrupted and resumed.