For future versions of XQuery, we should consider adding the ability to import a module and selectively replace a part of it.
There's a danger that adding something like this could destroy the ability to independently compile modules. Can you give an idea of scope and motivating use cases?
Motivating use cases are fairly clear I think: we're all familiar (from using object oriented languages and from XSLT experience) of how much can be achieved in terms of software reuse if you can use most of the functionality of an existing system while being able to change parts of its behaviour.
I think that concerns about separate compilation could be assuaged by (a) requiring functions that are overridable to be declared as such ("non-final"), and (b) requiring the overriding function to have a compatible type signature. (Both features that are sadly missing from XSLT).
Being able to independently compile XQuery modules is something that is incredibly useful in XQuery as it allows you to create systems that operate on Modules for example EXQuery's RestXQ and eXist-db's SOAP Server, both of these allow translation to/from XQuery function calls from a host system or language.
If the use-case is that being able to override aspects of a module helps re-use of code through composition, is there not an argument that the granularity of the users modules is not fine enough. i.e. they could just break a module into two modules.
I am not opposed to the capability or idea of overriding aspects of modules, as long as we can maintain the capability to independently compile modules somehow.
Assigning to future requirements per Working Group decision (https://lists.w3.org/Archives/Member/w3c-xsl-query/2012Oct/0087.html).