Peer review: Is this "citation tower" a bad practice?

I think this requires some nuance. You seem to suggest that citing the earliest usage is preferable, and it might be. But it might also be the case that each subsequent usage adds something to the context. That is to say that the citation appears in the chain for some particular use and the reader of the current paper might find it useful to explore that history. Each of those citations is surrounded by context that can be useful.

If you always cite only the earliest usage, then the reader has no hints about the overall usage context of that thing cited. And the current author might be inviting the exploration.

Of course, it might also be that the author(s) considers the lemma to be a minor point and has used the citation they found without chasing the chain back themself, not seeing any need for it in the current context.

Overdone, in mathematics, we just always cite Euclid.

So, I don't think citing a later, rather than an earlier paper is a bad practice necessarily, but could be in a particular instance.

You can always note the other citations in your review and point to the earliest one without making any specific recommendation.

In your case, the author of A that was also an author of F should have cited F instead of B. It's unlikely that that author forgot the reference F; it's more likely that they wanted more citations for B.

I recommend that you go with your option 4. Tell them to cite Z and F.

Vaguely calling something “bad practice” is, well, bad practice. You should be able to articulate precisely what’s bad about it. And it sounds to me like what’s bad about it is that it is misleading: it misleads the reader into thinking this lemma is new when it obviously has a long history, and that the credit for it is due to a different set of authors than the people it’s really due to.

With that in mind, I think everyone can agree that misleading readers is indeed “bad practice” and should be avoided, especially when it’s easy to do so. So it seems to me completely appropriate to ask in your review that the authors of “A” should cite the book “Z” either in place of, or in addition to, citing any subset of the papers “B” through “F”. And you can also suggest that they optionally add a brief comment outlining the history of the lemma and pointing out any relevant issues of notation as appropriate, depending on which of these sources they end up citing exactly.