Back to blog

What to do when a bulk product-photo run is interrupted

Handle a closed tab, uncertain submission or delayed output without blindly starting another paid job. Review saved state and quote only intended retries.

By CatalogCut4 min read

AI-assisted product documentation based on CatalogCut’s implemented workflow and the linked sources. Examples are illustrative, not customer case studies. How these guides are prepared.

A closed browser tab and a failed image job are different events. So are an uncertain submission, a slow result and an expired source. Treating all of them as “start again” can create duplicate work and an unnecessary new charge.

The first recovery step is to inspect saved state. Do not use a second request as a way to discover whether the first one was accepted.

When the tab closes after confirmation

CatalogCut catalog processing is durable after admission. Return to the saved run and refresh its state. Closing the page does not cancel the work that was already accepted.

Check which items were started, which are still waiting and which have completed. A run can contain different states at once. Do not infer that the entire remainder failed because one item is taking longer or the preview has not loaded.

Keep the run's identity in your team notes. Recreating the same selection under a different name is not the same as reopening it.

When you are not sure the confirmation arrived

A network interruption can happen after the server accepts work but before the browser receives the response. Use the application's saved-state and retry/replay controls rather than inventing another request or clearing a pending confirmation to make the page look clean.

CatalogCut uses request identity to distinguish an accepted replay from a different new request. That protection is not a reason to keep clicking every action: a new request with changed contents is a different decision.

If the state remains unclear, contact support with the run or item reference, approximate time and what the interface showed. Do not send passwords, authorization tokens or private storage links. CatalogCut support is the appropriate next step for unresolved account-specific outcomes.

Read the item state before choosing a retry

SituationSafer next step
The page disappearedReopen the saved run
The item is waiting or processingInspect saved progress; do not start a duplicate
A failure is confirmed and retry is eligibleReview the new retry quote
The image completed but you dislike itReview a separately quoted replacement
The provider outcome is unknownStop and reconcile; do not assume failure or refund
The source or result expiredCheck availability and select new evidence where needed

These distinctions also explain why a completed-but-rejected image should not be treated like a technical failure. The service may have successfully produced the paid result even though you choose not to publish it.

Do not hide an unknown outcome

An unknown provider outcome is not evidence that no work happened. Blindly retrying or refunding it can create inconsistent accounting or duplicate execution.

CatalogCut fences unknown catalog outcomes from ordinary retry, replacement and deletion paths that would hide the unresolved work. Wait for reconciliation or support rather than creating a workaround in another project. A dashboard that looks tidy is less important than a correct job and credit history.

This article explains the intended recovery flow; it is not a claim that every external-provider crash scenario has been independently tested.

Cancel remaining work deliberately

Use the run's cancellation control when you want to stop remaining work. Cancellation is not the same as retracting an already-dispatched external call, deleting successful outputs or promising that every reserved credit will be refunded immediately.

After cancellation, review the per-item states and totals. Keep any approved outputs you still need, subject to their individual availability. If the package is incomplete, explicitly choose whether a partial export is appropriate rather than treating it as a full delivery.

A useful recovery record ends with a clear answer for each item: approved, still running, confirmed failure, unresolved or unavailable. That is safer than restarting the catalog without knowing what the first run did.

Reopen the saved run after a lost confirmation response.

A fictional network interruption after the seller confirms processing.

  1. Find saved state

    Return to the saved run rather than creating a new selection to discover whether the first was accepted.

  2. Distinguish outcomes

    Separate waiting, completed, confirmed failure, unknown provider outcome and expired-source states.

  3. Choose intended recovery

    Use accepted replay or the appropriate quoted retry; ask support to reconcile unresolved outcomes.

Illustrative workflow. Review checkpoint: Closing a tab does not cancel admitted work; an unknown provider outcome must not be blindly retried or refunded.

Official sources