Algolia search cache invalidation limits for immediate index updates

When an Algolia index receives updates, the search API can serve responses from an edge CDN cache for a short window, and the JavaScript client may also retain responses in its in-memory cache for the session. The cache parameter is documented to control client-side behavior, but the interaction between index updates, CDN edge caching, and any request-level controls to force a fresh response is not clearly bounded in public documentation.

For applications that display near-real-time inventory or pricing, the combined effect of these layers can create a period where search results do not reflect the latest records. There is no widely documented API parameter or header that guarantees bypassing the CDN layer for a standard search query, and behavior may vary across infrastructure plans.

Is there a supported, documented mechanism to bypass Algolia's CDN edge cache for search requests, or is the only reliable approach to disable client-side caching and accept eventual consistency? If cache-control headers are honored at the edge, under which conditions and plan types?