Bulk email verification (up to 100 unique rows safe / 25 full)

Every email goes through the EXACT same verification path as POST /verify; this is a charged, capped, deadline-bounded loop over that path, NOT a separate/lighter verification. Bulk is NOT a rate-limit bypass: each row consumes the standard per-key/per-account /verify hourly budget (the SAME buckets single /verify draws from; singles and batches share one budget).

Input: exactly one of emails (JSON array), text (a paste blob: newline/comma/semicolon/whitespace-separated), or a multipart file upload (.csv/.txt, max 64KB). Tokens are trimmed, quote-stripped, capped at 320 chars, and exact-match deduplicated (order-preserving) before verification; N (unique count) is what gets rate-charged, metered, and returned as results. Safe mode caps at 100 unique rows; full mode (not generally available) caps at 25. N == 0 or N > cap is a whole-batch 422; a row that merely fails verification (bad syntax, disposable, a CSV header line) is never a batch-level error, only a per-row verdict.

Unlike single /verify (which consumes its rate-limit charge even on the request that gets refused), a batch's charge is CONDITIONAL: if the full N-sized charge does not fit the remaining hourly budget, NOTHING is consumed and the whole batch returns 429 (so retrying a large batch against a near-full budget never drains it).

The batch runs under a 60-second wall-clock deadline. Rows not started before the deadline return deliverable: "unknown", reason: "batch_budget_exhausted". The rate charge and usage meter still count them (no refund race).

Full mode applies the same checks as single /verify full mode: availability and plan once per batch (a 403 happens before any row work), rate limits and per-domain cooldowns per row.