Ok, I will re-format and change comment style to be consistent with the file.

I think you're right about not mixing C and C++ style comments unnecessarily, I used the C style comments since I copied from instructions which were supposedly working.
But checking the gcc info, C++ style should be accepted too, and maybe it will be treated better by clang-style.

Opened D227 and D228 for preliminary formatting as a basis for amending this one.

I didn't test this yet. But you wrote

"In practice it does not seem to work right despite -Wimplicit-fallthrough=3"

Note that you used "FALLTHROUGH" but on this page you linked, https://developers.redhat.com/blog/2017/03/10/wimplicit-fallthrough-in-gcc-7/:

it says "FALLTHRU".

Might that be the reason explaining this?

In D224#3770, @awemany wrote:

I didn't test this yet. But you wrote

"In practice it does not seem to work right despite -Wimplicit-fallthrough=3"

Note that you used "FALLTHROUGH" but on this page you linked, https://developers.redhat.com/blog/2017/03/10/wimplicit-fallthrough-in-gcc-7/:

it says "FALLTHRU".

Might that be the reason explaining this?

The Redhat page seems a bit older and only gives one example syntax (which I tried), but the gcc page lists a bunch of regexp's of which I tried a lot, in combination with different -Wimplicit-fallthrough=n , on gcc 7.1.0 .
I was surprised I couldn't get rid of the warnings - I checked with make V=1 that they were invoked on the command line, but maybe some other options overrides.
I guess it must be something silly like that.