How many are actually used? Nothing public carries call volume. It would take either registry telemetry or client-side data, and neither is published.
How many are maintained? The records carry publication metadata but this run did not capture last-updated timestamps per server. That is a straightforward extension of the method in the first note and would give a decent maintenance proxy — it was simply not measured here, and saying so is cheaper than implying it was.
Is the official registry representative? Several third-party directories exist with their own listings and their own inclusion rules. Overlap between them is unmeasured, so the official registry may be a large sample or a partial one, and there is currently no way to tell from outside.
Does the remote/local split hold by usage as well as by count? The reversal above is a count of published servers. It is entirely possible that local servers dominate actual invocations while remote ones dominate publishing, and nothing available would distinguish those two worlds.