No, again, I'm not "buying into marketing semantics". Just because you repeat that doesn't make it any more true, and it may just reveal your own blinkered outlook. And no, I'm not unable to explain, just unwilling to waste my time on those who can't, or won't, distinguish black from white or marketing hype from good practical techniques.
Doing a Save As in your example is just old-fashioned good practice when you open an original and create a derivative. Your question was so simplistic it's inane. What happens afterwards - to the derivative file - is what distinguishes what people mean by working non-destructively.
Traditionally you would reopen the file and do more work, something like transform or Silver Efex, for example. As your adjustment layers have masks (eg simulating a grad filter), you would need to get your pixels on a single layer. To avoid destroying anything, would you now apply your silly method of doing a Save As and Close, before continuing and flattening? Make another derivative file to meet your meaningless definition of non-destructive? You would? OMFG. Not very efficient, are we? But check on the big brain on you - in fact, you do have enough nous to create a new merged pixel layer, apply your Transform to this layer and close the file. The only way you can now fine tune the Transform or those adjustment layers is by discarding (ooh, destroying!) that new layer and repeating the previous steps.
The newer "non destructive" approach would be to convert those layers into a smart object, causing the Transform to apply as a smart filter. I still save and close, but my Transform edits remain fully editable. If I feel I need to edit the adjustment layers, I can do that. If I want to go back into the raw conversion, I can do that too. Infinitely-editable, non-destructive, good practice.
See the difference? Anyone else can.
John