Child document filenames are simply the given prefix plus a sequentially generated count. The child weaver is always the system html weaver.
Headings in child pages are adjusted to start at level 1. The 'print level' offset is not inherited from the parent or controllable in any way.
Heading numbers in child documents never reflect the position in the whole document, only the child page; that is, the first heading in a child spawned by 2.3.4 will be 1, not 188.8.131.52.
Each page will have it's own table of contents. There's no overall table for the document. (This will require links to pages other than 'self').
There are no 'next, previous, home' type links anywhere, these are necessary.
The spawning weaver assumes HTML. There's no 'in principle' reason it shouldn't be possible to generate other documents this way. Although, for example, 'hyperlinks' in printed documents as would be produced by LaTeX will be of the form, 'see page 99 of such and such', they're hyperlinks never the less, even if they have to be followed by the 'manual' browser. :-)
In light of these comments, one can say the current system confuses production of multi-page web documents, and production of distinct separate documents (which happen to be hyperlinked).