Feature request: Extend the transformation mechanism so it’s possible to rewrite numbers to their spellout form according to a language’s RBNF rules. This would be used to pre-process text before transcribing it to IPA; see also http://unicode.org/cldr/trac/ticket/8870 for further context.
Not sure how these rbnf transforms would know what language to use. Perhaps transform rules might need some way to pass parameters? Or perhaps it could be done with name hackery? For example, the IPA transforms for hy_AREVMDA (if we had them, this is just a hypothetical example) would call ::rbnf-hy_AREVMDA() at the beginning; the transform registry lookup would then instantiate the right rules on demand.