Suppressing the #tag expansion of inner content by hacking it out of CoreParserFunctions results in identical output and all Cite parser regression tests passing, and gives a sequential parse order which would simplify Cite in the ways we desire. Unfortunately, I don't see any existing mechanisms which would allow Cite to avoid parsing in this context.

I got curious about whether the inner content expansion can be skipped altogether, since the result will continue to be expanded in any case. Running the full parser test suite shows two new failures if we try this. Test cases are:

• Preprocessor precedence 11: found during visual diff testing
• Parsoid: unknown parser function (T314524)

But the way these fail could be more of a technicality and not something that editors rely on... maybe?

Okay it's not quite that easy. The inner content does need to be expanded when it's transformed into XML-ified wikitext but the outer tag isn't handled by an extension. In that case, {{#tag:b|This is a known parser function}} becomes <b><part><name index="1"/><value>This is a known parser function</value></part> </b> for example.