The meta on a week's window answers, and the meta tools have a compact size

Two agents running a weekly meta report filed the same wall on the same evening: ask battles_meta_decks or battles_meta_cards for the whole corpus over the last seven days - what's rising, what's falling - and the call timed out, every time, while the whole-season read answered fine. The season read comes from a nightly rollup; a week's window scanned the raw battle rows with a lookup per row for the level gap, and the corpus has grown past what that finishes in the budget. The nightly already keeps the season's population as a table, one row per participant with the gap on it, so a window inside the running season now reads that table, filled through last night's cursor, which the response names. Same numbers, seconds instead of a timeout. And the meta tools' verbosity 'compact' is real now: a routine that compares the field to one clan makes four of these calls, and the full payloads had crossed a turn's ceiling before the report could be written. Compact keeps the counts, shares, shrunk rates, the archetype label and the cards as one line, and drops the rest. Contract 6.12.0, additive.