A lot of comments here about observability. Well UKL solves this nicely because you can run regular userspace processes in the unikernel, thus you can (optionally of course) run a shell, perf, and anything else you need to observe what the machine is doing.
Linux is about 40 million lines of code, and even if you disable almost everything optional, it's hard to get a usefully bootable Linux below about a megabyte.
The seL4 kernel is about 9000 lines of code, a code base that's 4000 times smaller, and I think it compiles to about 80KiB, about 12× smaller than Linux.
I elicited a small unikernel web server from Claude the week before last, and, not counting lwIP, it's 385 lines of code. The built unikernel image, which does include lwIP, is 58K, about 17× smaller than Linux.
The KokalOS kernel I've been working on over the last week, originally by refactoring the unikernel web server, is currently about 900 lines of code (65536× smaller than Linux) and compiles to just under 10KiB (128× smaller than Linux), but it's currently just running a CLI and a web server with a single Ethernet card. I anticipate that in a finished state it will probably be close to the size of seL4.
Sel4 is cool stuff.
Still, there's a whole pile of stuff there that many applications don't need. If you truly can run the entire of the OCaml runtime and standard libraries etc without it, and you're happy writing your apps in OCaml... MirageOS has always felt winning-ish to me as a general idea.
And to TFA's point, now with LLMs the relative exotic nature of OCaml as a platform need not be a huge hindrance. I like the language, but have only dipped my toes in. Definitely tempting to try writing apps on it w/ MirageOS now. Though I doubt I'd convince anybody to pay me to do that...