The algorithms that can recover data after the corruption like bzip2 and lzip don’t see much use these days
- if you know tar supports your directory / metadata, you don't have to lookup if zst will preserve that info
- interfile compression
I don't think people would say they use tar because it's a stream format. They use it because it packages files/directories into a file
If I expect I may have to deal with some form of data corruption (such as long-term archival media), I'd rather add a layer such as PAR2 [0]. Rather than content myself with "if the tar file is corrupted, I can probably recover most of the contents", erasure coding gives me confidence that I can definitely recover all of the contents, with confidence they're not corrupted, as long as I have enough parity data.
windows-only, though (and 3rd-party)...
The fact .tar.zst or .tzst isn't supercommon (yet) is for the same reason .tar.xz took years to become widely adopted: lots of OSes still ship `tar`s that don't support it out of the box, like Windows (which only supports gzip actually) and macOS (which ships an old version of bsdtar and no libzst). On a reasonably new Linux, is like opening every other kind of tar.