Glad to hear it wasn’t all quite as messy as you remembered it being. It’s funny in a way because Scrivener has had two particular tools available to it, the auto-save and snapshot features, long before the concept was terribly popular—so it does do some things differently. With macOS 10.7, Apple introduced a similar mechanism system-wide, but with slightly less control over when a version is made. I find their implementation, although it doesn’t require any awareness to be using it, rather too far over on the side of chaff. If I browse through versions for documents I see micro-edits being preserved as whole slices in the version stack, and finding what I want can mean trawling through dozens of sentence level edits. And well, the less said about the Star Trek interface that blots out the entire computer until you are done with it, the better. 
I do prefer the more intentional mechanism Scrivener uses. To me, ⌘5 is a bit like ⌘S (and not just in appearance!). I use it whenever I’m about to change course while editing, and I use it when I’m done. I can browse “versions” that were made for human reasons in a comfortable and integrated environment. Make ⌘5 as habitual as saving in Word, or ⇧⌘5 as habitual as using Save As, and you’ll rarely run into cases where a sentence gets lost to an over-aggressive edit.
To summarise our offerings in terms of safety nets:
- Fully automatic (by default): project backups. Every time you close the whole thing gets backed up, down to label changes and compile settings, let alone text.
- Automatic snapshots (as an option): triggered whenever you save the whole project.
- Manual backups: setting aside simply triggering the automatic backup system, manually setting aside a stack of backups in another location provides that same level of human reasoning for having them. The automatic stuff is good—like Time Machine and Versions can be—but a stack of files that isn’t rotated by software and that can be labelled meaningfully is often a life saver. If you accidentally modify an old book that is laying around in the binder for reference, you can pull up the milestone that was created when that book was finalised as a project and drag in as much binder tree as needed to restore it intact.
- And of course by that same token, manually created snapshots fall into that same realm. You can have a somewhat macOS Versionesque spool of incidental time frames, but unlike macOS you can have labelled milestones thrown into the mix too.
For the stuff we write that we probably won’t want to edit again, that’s all well and good, but for the stuff that we definitely don’t want to edit, ever, try this:
- Hit ⌘P from the editor on the text you wish to preserve.
- From the print preview, use the “PDF” button to send a copy of the file back to Scrivener.
- Maybe copy and paste the original text it came from into the Document Notes for the PDF.
PDF remains searchable, can be annotated simply in Scrivener or more thoroughly in dedicated readers, and is pretty good for copy and paste (though the sidebar Notes copy would probably be preferable for that). And of course a similar approach can be adopted from the start—instead of dragging a .docx into the binder, just open it in LibreOffice or whatever and print it to Scrivener. If something truly is never meant to be touched, why import it as a text file at all?