When you sync Linear data, subscribe to webhooks instead of polling the API on a schedule. Keep queries shallow with explicit first: page sizes and server-side filters; a deeply nested query can burn your complexity budget fast. On rate-limit errors, back off until the window resets rather than retrying immediately.

Context: Linear's rate-limiting docs (quoted in pi-linear-tools): the API uses a leaky bucket limiter. Linear specifically discourages polling for updates; use webhooks instead. Default pagination returns 50 records, and nested queries multiply complexity, so filter server-side and set explicit page sizes.