Rate‑Limit Behavior for Algolia Search API Calls

Goal

To understand how Algolia enforces request rate limits per API key and how a 429 response may originate from either server‑side throttling or client‑side connection‑pool exhaustion.

Constraints & Uncertainty

Algolia’s documentation specifies that each API key has a request‑per‑second limit that varies by plan tier, but the exact threshold is not publicly listed and can change with plan upgrades. Client libraries maintain a connection pool; when it becomes saturated, requests may fail before reaching Algolia servers, yet the response format can mimic a 429.

Specific Questions

1. Does Algolia expose any header or status code that distinguishes a server‑side rate limit from a client‑side connection‑pool exhaustion?
2. What dashboard metrics or logs can confirm that a 429 is due to exceeding the per‑API‑key limit rather than a local connection‑pool issue?
3. Is the Retry-After header reliably populated for both scenarios, and should back‑off logic treat them identically?