On 27 Dec 2015, at 21:57, Wolfgang Schuster <schuster.wolfgang@gmail.com> wrote:

Meer, Hans van der
27. Dezember 2015 um 21:47
One more observation on the strangeness of \setuppagenumber.

Putting \setuppagenumber[number=4] before the call to \xmlprocessbuffer and putting \setuppagenumber[state=stop,number=8] inside the \startxmlsetups does make the pagenumber having the value 8. This proves two things:

(1) the \setuppagenumber in the \startxmlsetups is processed before the pagenumber is fixed and typeset

(2) the [state=stop] in the \startxmlsetups is wrongfully ignored, whereas at the same time [number=8] does get honored.

Imho this demonstrates there is an error in the processing of \setuppagenumber when located inside the \startxmlsetups. An error that does not occur when called at the \starttext document-level before the \xmlprocessbuffer. 
Could it be a matter of locality?

You have to provide a *working* minimal example.

Not so simple, alas. 
But I will be greatly helped if you can ascertain that

\startxmlsetups
code within
\stopxmlsetups

is code executed locally as if between \begingroup .. \endgroup.
That would explain the behaviour.
The remedy then could be a more global execution of \setuppagenumber.
I already tried in vain \global\settuppagenumber
Hans van der Meer