Edge Cache Consistency After Switching From Build-Time Deploys to On-Demand Revalidation

Answer

A browser request for a static asset immediately after an on-demand revalidation trigger can return the pre-deploy representation. There is no universal instant visibility guarantee for Jamstack hosts. Immediate visibility is provider and plan specific and is controlled by CDN purge/invalidation behavior, TTL, stale-while-revalidate settings, and browser cache headers, not by revalidation completion alone.

Confirmed facts

• Jamstack is an architecture, not a versioned product. No universal consistency contract exists across providers for on-demand revalidation.
• Moving from build-time publishing to on-demand regeneration shifts the consistency boundary from build completion to revalidation completion.
• Edge propagation remains eventual. It depends on the CDN, plan, TTL, stale-while-revalidate settings, and browser cache headers.
• Idempotency keys for retries belong to the backend API, not the static frontend cache.

Likely explanation

• A revalidation success does not equal global cache purge completion. Edge nodes may still serve a cached copy until TTL expires or purge propagates.
• Instant purge or cache invalidation APIs, where offered, are provider and plan specific and may be separate from on-demand revalidation. Purge does not automatically guarantee immediate post-deploy visibility for all assets.
• Browser cache headers can prolong stale responses beyond edge propagation, so a fresh browser session may still see old content.

Steps needed for this case

1. Confirm provider name and plan. Review documented revalidation latency, purge limits, and whether purge is included with on-demand revalidation.
2. Trigger revalidation and inspect response headers from multiple locations with a fresh browser session: Age, Cache-Control, CDN cache status.
3. Compare asset ETag or content hash before and after revalidation to detect stale delivery during propagation.
4. For critical assets, use versioned URLs or short TTLs and avoid assuming immediate consistency during the propagation window.

One missing diagnostic detail changes the recommendation: which provider and plan you are on. Provider and plan determine whether instant purge applies to on-demand revalidation or only to full builds, and what latency guarantees are documented.