... In my view the "dominant use"-arguments are fallacious, since they
only relate to quantity (pragmatics) and not to quality (semantics). ...
In my view your "uncancellable" arguments are fallacious, because they
imply a static view of language, as something which obeys certain rules
without exceptions, which simply does not apply to real language. The
consequence of applying your method to real text corpora of any size, in
any language, is that exceptions are found to every proposed rule, and
therefore nothing is uncancellable and so by your argument nothing is
semantic, everything is pragmatics. Now maybe there is something in that
final conclusion. But you need to realise that your arguments that
WAYYIQTOL is not a semantically distinct verb form can be applied to
pretty much any verb form in any language, at least for which there is
adequate evidence.
Even your example "the English verbs "went" and "spoke" have an
intrinsic past tense and the forms are grammaticalised location in time"
is by no means absolute, for you are sure to be able to find a minority
of uses (even if they are considered non-standard by prescriptive
grammarians, and quite apart from the special cases which you mention)
in which they are used with non-past meanings. And by the way, in
hypothetical conditional clauses like "If he went to the theatre this
evening, he would see..." (spoken of someone who has made a definite
decision not to go), there is no past tense meaning component to "went"
and so the past reference is entirely cancelled. So, by your arguments,
the English past "tense" is entirely pragmatic.