The only thing I can think of is to clone the wiki, and for WikiA, delete PageB, and for WikiB, delete PageA. This won't remove the histories of those pages, but it will at least remove those pages.

If just deleting the unwanted pages is insufficient (e.g. perhaps they contain information that must not be made public), you can split a git repository (including ikiwiki repositories) with git filter-branch (see this stackoverflow question, for instance).

This preserves the history of each individual page, but rewrites the entire history of the repository (it re-does every commit, pretending that the addition of the omitted pages and every subsequent edit to them had never happened); it's like git rebase but more so.

As a result, existing branches will no longer be able to push to the rewritten repository.

@smcv: Thanks, that looks promising. The example from stackoverflow is with subdirectories. What do I have to change to match a list of single files (pages) instead?

I have put the question on stackoverflow: http://stackoverflow.com/questions/8443372/split-an-ikiwiki