Johannes Schindelin [off-list ref] wrote:

As I just pointed out in
https://lore.kernel.org/git/nycvar.QRO.7.76.6.2202251600210.11118@tvgsbejv=
aqbjf.bet/,
it seems that replacing the static check presented in v1 by a runtime
check needs to be reverted.

In addition to the example I presented, there is another compelling reason
to do so: with the static check, we can detect incorrect usage strings in
all code, even in code that is platform-dependent (such as in
`fsmonitor--daemon`).

First of all, thank you so much for putting so much time to look into
my PR. I appriciate your research about various possible outcomes of this
Patch request.

I saw your mail where you listed some of the disadvantages of the current
version. I also agree with the arguments you provided and it is also true
that one wouldn't find any clue by seeing the output of the `CI` link
you mentioned (i.e. https://github.com/git/git/runs/5312914410?check_suite_focus=true).

Let's see what Junio, Ævar and others say about this.

Thanks :)