What happens to in-progress orders if a payment channel temporarily goes down?
An unavailable payment page, an API timeout and a customer charged while an order remains unpaid may be three different faults. During a temporary channel outage, the most dangerous response is marking all in-progress orders failed and asking customers to pay again. 稳付pay explains which order states to preserve, which results to investigate and how recovery prevents duplicate charges and fulfillment.
Contents
First distinguish where the outage occurred and whether the payment result is known
Payment requests, channel processing, responses, callbacks and merchant fulfillment are separate stages. A request may reach the channel and complete while its response is lost, or the channel may work while the merchant's callback endpoint is unavailable. A page error or API timeout first means this communication produced no reliable result, not proof that payment never happened.
Adyen's official error-handling documentation distinguishes cases such as request timeouts and recommends checking transaction status when no result is received. This is a general design reference, not a claim that 稳付pay uses Adyen APIs or timeout settings. Establish payment results through trusted original-channel queries, verified notifications and reconciliation evidence where necessary.
Preserve facts in state records; timeouts must not overwrite confirmed results
| Example state | Known fact | Handling during an outage |
|---|---|---|
| created: order created | A local order exists without a valid payment result | Pause submission and show temporary unavailability; do not assume payment. |
| pending: awaiting payment or processing | A payment attempt exists; its result is not yet confirmed | Preserve the attempt, receive callbacks and query status as specified. |
| paid: payment confirmed | Trusted evidence of success has been obtained | Preserve payment; API unavailability must not revert it to unpaid. |
| fulfilling: fulfillment underway | Shipping, activation or code delivery has started | Inspect fulfillment jobs and retry delivery, not collection. |
| fulfilled: fulfillment complete | The corresponding entitlement has been issued | Retain delivery records; duplicate notifications must not issue again. |
| expired: locally expired | The merchant's payment window has ended | Check whether the original channel closed it; handle late payment separately. |
| unknown: result needs verification | Current evidence proves neither success nor failure | Queue for reconciliation and avoid prompting another payment. |
These are illustrative business labels, not universal API enums. Store payment, fulfillment and settlement states separately. unknown may be an investigation flag but must not erase a previously confirmed paid state. Refunds, reversals and disputes are separate events, not replacements for the payment record through an unpaid status.
Pause or reroute new orders while preserving the original channel for existing ones
Once an outage is confirmed, stop submitting new payments to the affected channel. Eligible new orders may use a validated backup. A backup does not mean editing the channel ID of an existing order; its original transaction may still complete.
Bind each attempt to the business order ID, channel identifier, channel transaction ID, merchant account, amount, currency and creation time. Even after new orders switch channels, retain the original callback verification configuration, query path and mappings until those transactions are reconciled. Otherwise, a late success may have no identifiable order to update.
Before changing payment methods, use the original channel's capabilities to establish whether the prior attempt is canceled, closed or definitely unpaid. Creating another payment with an unknown prior result may succeed on both channels. Cross-channel idempotency is not automatic; add verification and duplicate-payment handling.
Retry the same operation instead of repeatedly creating new payments
Stripe's idempotent request documentation explains identifying repeated requests with the same idempotency key. Generally, persist the payment attempt and request identifier before first submission, then reuse that identifier and identical parameters under the channel's retry rules after network errors, instead of creating a fresh transaction at every timeout.
Local uniqueness constraints and concurrency control must also prevent multiple workers submitting the same operation. Whether an order ID is the key, its lifetime and which errors permit retries depend on channel documentation. Adding a custom request field does not automatically give the upstream service idempotency support.
Use bounded retries with backoff during recovery so accumulated requests do not overwhelm the channel. For confirmed payments, advance fulfillment or notification compensation only; do not create another payment to “refresh status.” Different amounts, refunds and new attempts need separate identifiers with retained relationships.
Verify and persist callbacks before advancing business processing
Webhooks may repeat, arrive late or interleave with query results. Verify signatures, request bytes and timing as required by the channel, then check account, order, amount, currency and event type. Source addresses can be supplemental controls but cannot replace signatures and business-field validation.
Write valid events to a persistent inbox or reliable queue, and acknowledge receipt only after commit. Slow fulfillment and notification work belongs in subsequent jobs. Returning success after only an in-memory write can stop upstream redelivery while leaving no recoverable work after a crash.
Deduplicate by channel event identity and enforce a second unique constraint for business fulfillment. One payment can generate multiple distinct events, so deduplicating HTTP requests alone is insufficient. Stripe's official Webhook documentation also discusses duplicates and delivery ordering. A late pending status must not overwrite confirmed success, and rerunning a fulfillment job must not issue another code.
Recover through three evidence paths: redelivery, status queries and reconciliation
- Callback redeliveryRestore the endpoint and replay valid notifications not successfully handled.
- Active status queriesVerify unresolved payments with the original channel and transaction identifiers.
- Difference reconciliationCompare channel, local payment and fulfillment records.
Queries may also be rate-limited or unavailable. Schedule batches and retain the latest trusted result. A successful query request does not mean a successful payment: read the business outcome. Missing or conflicting content remains awaiting verification, not an invitation to guess.
Reconciliation finds differences such as channel-only transactions and local paid orders still unfulfilled. Adyen's reconciliation documentation illustrates transaction and batch reconciliation; this site does not copy its report fields. Payment confirmation, customer fulfillment and merchant settlement require separate checks. Real-time USDT settlement does not replace order-state recovery.
Handle post-expiry payments and refunds separately
An expired merchant-page countdown only ends the local payment window; it may not cancel the channel transaction. For late callbacks, distinguish an on-time payment reported late from a payment made after expiry. Preserve payment and notification times rather than discarding all late messages or unconditionally restoring released stock.
If a valid late payment can still be fulfilled, restore the order under previously explained rules. Otherwise, use manual reconciliation or the agreed refund process. Record requested, processing and completed refund states separately. A submitted request is not proof of receipt; use the channel's confirmation and receipt rules.
During the outage, preserve confirmed payments and their evidence. Record actual refunds or reversals as new events and amounts, rather than deleting payment history. This supports both customer explanations and later reconstruction during reconciliation.
Two scenarios: recover the order instead of collecting payment again
Scenario one: a QR payment was charged, but the merchant page keeps waiting
The upstream payment completed while the merchant's callback endpoint was briefly unavailable. Show verification in progress and retain the original query and notification mappings. After recovery, establish paid through valid redelivery or a status query, then fulfill once. Closing the order and asking for another scan may make both attempts succeed and increase duplicate-payment support costs.
Scenario two: a paid order's code-delivery job times out, then another notification arrives
The system recorded paid, but the fulfillment worker did not report promptly after execution. Recovery first checks delivery records or reserved-code relationships before reissuing. The second notification completes state handling without consuming another code or creating payment. If code exposure cannot be confirmed, investigate manually rather than assuming a timeout means nothing was sent. These are instructional scenarios, not incident reports.
Triage by severity and tell customers the actual progress
- Level one, entry failure: inspect the payment page, DNS, certificates and API responses, determine whether only new orders are affected, and pause the affected entry point.
- Level two, unknown result: collect unresolved attempts, verify original channel identifiers, limit duplicate submissions and schedule status checks.
- Level three, paid but unfulfilled: inspect persistent events, queue backlog and fulfillment records; restore entitlements first while preserving idempotency.
- Level four, records do not reconcile: investigate late and duplicate payments and refunds, preserve evidence and assign an owner.
Customer wording may say: “We are verifying your payment result. Please do not pay again; we will continue handling the original order.” Once confirmed: “Payment is confirmed and your product is being processed. No additional payment is needed.” For new orders only: “This method is temporarily unavailable. Choose an available method; for submitted orders, verify the original payment first.”
Explain the scope, known status and follow-up channel without promising an unverified recovery time. The 稳付pay merchant backend remains under development; use the actual contact details for integration or exception reconciliation. Queries, redelivery and retries described here are design recommendations, not claims of available buttons or APIs.
References and technical documentation
The official sources below support protocol and implementation checks. Code and scenarios illustrate technical concepts; validate your environment and production integration protocol separately.
Keep improving your site
Check each step of networking, orders and funds notifications against your business scenario. Discuss payment integration and USDT monitoring through the contact page.
Contact 稳付pay ↗