X removed the post-count caption from each cell on `/explore/tabs/trending`.
The adapter still iterated `divs[2..]` looking for a numeric text node and
fell back to the literal string "N/A" when it found none — which was every
row, on every call. We were emitting a silent-wrong column for every result.
Drop the column and the no-longer-relevant scan loop. Add a regression test
on the columns shape so the column doesn't slip back in.
Live runs of `opencli twitter trending` previously returned rows like
`{rank: 1, topic: "...", tweets: "N/A", category: "..."}` — the `N/A` was
not a transient outage, it was structural.
jakevin
aa8d4b72f7