Most of these are not really serious issues in the real world. It's been critiqued extensively in the past (posted those discussions in a separate comment) and I'll not repeat it here. But I will add that lzip's "yo, here's a tarball, trust me bro"-model of distribution is no longer fit for 2025, for reasons that should be obvious. In my opinion this blows any concerns about xz out of the water. Technically this is about formats and not implementations, both de-facto have one dominant implementation, so that's a bit of an academic distinction.

Anyway, zstd is the future for most use cases.

I'm indeed interested in the author's opinion about the zstandard format. Because one seems to use xz as a kind of scapegoat...