Not very important, but at least it should be a (minor) improvement.

> A test suite doesn't have to
> assume that the internals work as they should, it should just test
> that.  So testing both sounds to me better than testing just one
> assuming that this one covers both.

I agree, we should definitely test raw bytes inserted from both unibyte and multibyte strings. The current code doesn't do that, hence my patch.

With respect to other values there should be a reasonable qualitative difference between test cases: testing both #x80 and #xfc isn't more useful than testing the display of both the letters 'A' and 'B'. But if you think that the current code is deficient in coverage, please do propose extensions and I'll adapt the patch accordingly.