Keeping Paywalls and Purchases Alive During an Outage
A subscription platform sits inside its customers' revenue path, so an outage is not just an internal incident. This guide describes a degraded-mode design that can keep paywalls visible and purchases flowing while the main service is unavailable, then reconcile deferred work once service returns.
Why a purchase path needs a degraded mode
When the subscription backend is a critical dependency, its availability shapes the app's reputation and the end-user experience. The original design already tolerated some connectivity problems through caching and retries, but the purchase flow still depended on the backend being reachable to unlock entitlements. A run of disruptions, some from cloud providers and some internal, made that gap visible.
The goal of the degraded mode is narrow and practical: keep the purchase experience usable while the primary service is down, and lose no data. It is not a full replacement for the live system. It is a limited mode that preserves key functionality, including when the underlying store services are themselves disrupted.
What offline entitlements did not cover
Offline entitlements already handled the case where servers could not be reached. Two scenarios remained open. First, apps on older SDK versions still had problems during API outages. Second, an app needed server data to know which products to show in the paywall and which entitlements a completed purchase should unlock. A customer who installed before the outage was usually fine because the data was cached; a customer who installed during the outage had no cached data to fall back on.
Two mechanisms: cached display and deferred replay
The design separates two jobs that are easy to confuse.
When the API is down, the paywall reads products from a snapshot file that is refreshed once a day and mirrors the live API response, so the paywall can still load. This is a read path: the app gets a recent snapshot instead of a live answer.
Purchase handling is a write path with a delay. Normally the app sends a transaction to the backend, which verifies it, identifies the unlocked entitlements, and returns them. In degraded mode the entitlements are unlocked automatically for a limited period. The actual HTTP requests are logged and replayed against the servers after the primary service recovers, so the transaction is not lost.
What degraded mode deliberately skips
The mode is honest about its limits, and those limits matter when you decide whether to rely on it.
- While the mode is on, experiments and any per-customer offering customization are bypassed.
- Outbound webhooks and events to third-party integrations are paused during the outage; they are queued and delivered once the logged requests are replayed after recovery.
- App-store validation of purchases is deferred while the mode is active.
These are trade-offs, not oversights: the system prioritizes continuity and no data loss over immediate validation and personalization.
What the incident record shows
The following examples are historical references from the operator's own incident record. They describe specific dates and conditions, not a guarantee of current or future performance. Check the current system status and documentation before relying on any figure.
- On May 7, 2026, the mode was enabled after an API capacity loss. For about 37 minutes it issued temporary entitlement grants, and all buffered transactions were replayed afterward with no transactions lost.
- On October 20, 2025, a major AWS us-east-1 outage caused widespread API degradation. The mode was activated but constrained to roughly 20% capacity because the outage was regional, and all purchases were replayed after recovery.
- On March 29, 2023, about 450,000 purchases succeeded during a disruption that affected one storage layer.
- On April 13, 2023, the mode helped 13,000 people use an app they had just purchased while Apple itself was having trouble processing new purchases.
A checklist before you rely on degraded mode
Use this as a decision aid, not a substitute for the current product documentation.
- Confirm which SDK versions your app fleet runs. Older versions were the original gap.
- Decide whether a daily-refreshed paywall snapshot is acceptable for your catalog and pricing changes.
- Confirm that ignoring experiments and per-customer offering customization during an outage is acceptable for your business.
- Confirm that delayed third-party webhooks and events are tolerable for your integrations.
- Confirm that deferring app-store validation until replay fits your fraud and refund posture.
- Confirm that a limited-period entitlement grant matches your entitlement and access rules.
- Verify that the replay step is monitored and that buffered transactions are reconciled after recovery.
Worked example: the May 7, 2026 window
Suppose the API loses capacity at 10:00. The mode turns on, and for roughly 37 minutes new purchases unlock entitlements through temporary grants rather than live verification. During that window, the paywall is served from the cached snapshot, experiments and per-customer customization are ignored, and third-party webhooks are held. Each transaction's HTTP request is logged. When the API recovers, the logged requests are replayed, the held webhooks and events are sent, and the entitlements are reconciled against the verified transactions. The outcome recorded for that window was no lost transactions. The lesson is that the recovery step, not the outage step, is where correctness is finally established.
Where the design is heading
The limitations above are the active work. The direction is to keep the continuity benefit while closing the gaps around experiments, integrations, and app-store validation. For teams evaluating this pattern, the useful question is not whether an outage can be eliminated, but which functions must survive one and which can safely wait for replay.
Frequently asked questions
What is the Subscription Platform Fortress degraded mode?
It is a limited degraded mode that keeps paywalls visible and purchases flowing while the primary service is unavailable, then replays the logged requests after recovery so no transaction data is lost.
How is it different from offline entitlements?
Offline entitlements handle the case where servers cannot be reached. The degraded mode also covers older SDK versions and the case where an app is installed during an outage and has no cached product or entitlement data.
What is skipped while the mode is enabled?
Experiments and per-customer offering customization are ignored, third-party webhooks and events are held until replay, and purchases are not validated with the app stores during the outage.
Where do paywall products come from during an outage?
From a cached file storage that is refreshed daily and mirrors the data the live API would return.
How are purchases reconciled after recovery?
The actual HTTP requests are logged during the outage and replayed against the servers once the primary service recovers, and the held webhooks and events are sent at that point.