Delete many sandboxes, one array-shaped call - the teardown shape a loop wants [rate class: sandbox_api_rps]

Delete sandboxes and stop their RAM-minute meters. SCOPE: microVMs you forked in your own workspace - disposable by design and re-forkable from their snapshot at any time, so a delete is the routine close of a fork's work, never a loss. Array-shaped, like sandbox_exec: pass sandboxes: [...] and tear down a whole batch in ONE call. A loop that deletes N forks one at a time pays N round trips for work that is one round trip; the array shape is not an optimization, it is the shape. Sandboxes bill for as long as they exist, so delete each fork the moment its work is scored - forks are disposable by design, because anything worth keeping should be a snapshot (snapshot_sandbox), not an idle sandbox. Deleting a sandbox never touches the snapshot it was forked from: re-fork an identical copy any time. Results come back in request order, one per entry, each carrying deleted: true|false - and a false entry is a sandbox that is STILL RUNNING AND STILL BILLING, either because the id was not yours or because the host refused the teardown. The call still succeeds when that happens, so check the entries, not just the status: deleted less than the number you sent means read them.