Decide When a Campaign Queue Is Stalled Rather Than Merely Slow
Prepared with AI assistance by Alactic Editorial. Examples identified as hypothetical are illustrative, not customer results.
A campaign still says queued, its pending count barely changes, and someone asks whether to submit it again. Do not use elapsed time alone to decide. Treat a queue as suspected stalled when comparable observations show no meaningful forward movement, or when older work remains stranded while other work advances. Before authorising recovery, establish what happened to the pending messages and whether they are still appropriate to send.
The immediate choice is not simply wait or retry. You can continue observing, investigate a specific uncertainty, or pause further submissions while preserving evidence. A campaign can also be progressing technically but moving too slowly for its message to remain useful.
The worksheet below turns those distinctions into an operator decision. It is a recommended operating practice, not a description of controls available in a particular product.
1. Establish what queued means in this case
Start with the state you can actually observe. Campaign preparation, template review, submission for processing, and delivery are different stages. A campaign waiting for preparation or review is not evidence that a delivery queue has stopped.
Record the workspace, campaign identifier, submission time, displayed state, and the source of that information. Then ask what the displayed state represents. Does it describe the whole campaign, individual messages, or only the latest summary? If that meaning is unknown, record the uncertainty rather than interpreting queued as proof that nothing has happened.
Use a fixed scope for the investigation. A campaign-wide total can hide one stranded segment; a workspace-wide total can move because an unrelated campaign is processing. Wherever evidence permits, follow the same campaign and the same pending cohort across observations.
- Pending: Work still awaiting a known next step, according to the available status definition.
- Processing evidence: A recorded attempt, handoff, acknowledgement, or other event that identifies which stage advanced.
- Outcome evidence: A delivery result or explicit failure, rather than an assumption based on disappearance from the pending list.
- Unknown: Work whose current disposition cannot be established.
Do not turn unknown into failed merely to make the totals reconcile. That shortcut can lead to duplicate sends during recovery.
2. Observe movement, not just the pending total
A pending count is useful only with context. It may stay constant because new work is entering as old work leaves. It may fall because messages failed or were cancelled, not because they were delivered. It may also reflect an old summary rather than current processing.
At each observation, capture the timestamp and timezone, pending count, oldest pending item and its age, latest processing evidence, new submissions since the previous observation, and any changed outcomes. If individual message evidence is unavailable, state that limitation explicitly.
The oldest pending work is particularly useful. A falling total with the same oldest item can indicate partial progress rather than a fully healthy queue. Conversely, an increasing oldest age is not enough by itself to prove a stall: older items may still be advancing through a later stage, or the displayed data may be delayed.
Choose an observation interval that fits the campaign's relevance deadline and the refresh behaviour of the evidence source. Do not invent a universal timeout. A documented processing expectation, if one applies, can inform escalation; an operator's preferred waiting time is not a provider guarantee.
When possible, avoid adding new work to the observed scope during diagnosis. If submissions must continue, record them separately so their arrival does not conceal departures.
3. Filled worksheet: a campaign that moves, then stops
Fictional worked example: Every identifier, timestamp, count, and deadline in this worksheet is hypothetical. The observations illustrate a decision process, not an Alactic performance benchmark or promised processing time.
In this example, an operator is monitoring a campaign about an afternoon collection slot. No new recipients are added after submission. The campaign owner has set a relevance cutoff after which the unchanged message should not be sent.
QUEUE-OBSERVATION WORKSHEET — HYPOTHETICAL Scope: Campaign COLLECTION-A, workspace Operations Observation date: Fictional operating day; all times UTC Submitted: 09:00 Initial submitted cohort: 600 messages Message relevance cutoff: 12:00, set by campaign owner New work added after submission: None Evidence available: Campaign summary and message event export Evidence limitation: No visibility into internal processing workers Owner: Campaign operator Recovery authoriser: Campaign owner Snapshot A — 09:10 Pending: 600 Oldest pending: Item P001, submitted 09:00; age 10 minutes Progress evidence: Submission recorded; no later event visible Investigation trigger: None established from this snapshot alone Action: Observe again at 09:20; do not resubmit Snapshot B — 09:20 Pending: 480 Oldest pending: Item P121, submitted 09:00; age 20 minutes Progress evidence: 120 items now have downstream acceptance events Outcome evidence: Delivery results not yet established for all 120 Interpretation: Processing movement exists; delivery not proven Action: Continue observing; next check 09:30 Snapshot C — 09:30 Pending: 480 Oldest pending: P121; age 30 minutes Progress evidence: No newer processing events since 09:18 Summary checked: Fresh export agrees with displayed pending count Investigation trigger: First unchanged interval after observed movement Action: Inspect available errors and notices; next check 09:40 Snapshot D — 09:40 Pending: 480 Oldest pending: P121; age 40 minutes Progress evidence: Same cohort pending; no newer processing events Available errors/notices: No explanatory event visible Investigation trigger: Repeated unchanged cohort, count and event time Chosen action: Escalate as suspected stall; pause related new submissions Do not: Retry the full campaign or classify unknown items as failed Next review: 10:00, assigned to campaign operator At escalation Known movement: 120 downstream acceptances Unresolved work: 480 pending; downstream outcomes checked separately Question for support: Is remaining work waiting, blocked, or already handed off? Recovery condition: Establish disposition and authorised recovery scope Relevance condition: No unchanged release at or after 12:00
The decision at Snapshot D is deliberately limited. The operator has evidence for a suspected stall, not proof of its technical cause. Pausing related submissions protects the investigation from additional uncertainty; it does not imply that already-submitted work has stopped.
The earlier movement also matters. It narrows the question from whether the campaign ever entered processing to why the remaining cohort has not advanced since the last observed event.
4. Use evidence to choose between waiting and investigating
Waiting is reasonable when fresh evidence shows the same pending cohort advancing, the message remains relevant, and no unresolved condition requires intervention. Record the next check and its owner. Open-ended waiting is not a diagnosis.
Investigate when an observation contradicts that picture. Useful triggers include:
- The same pending cohort and oldest item remain unchanged across repeated, fresh observations.
- Newer work progresses while a particular older cohort remains behind.
- The summary and underlying records disagree about what is pending.
- A processing error, restriction, or dependency notice appears.
- The latest meaningful event becomes increasingly old as the relevance cutoff approaches.
Check the evidence source before escalating a processing fault. Refresh or obtain another authorised view where available, and compare its observation time with the event time. A recently opened screen can still contain old data.
If available error records show HTTP 429, that indicates rate limiting. A Retry-After header may specify when to retry, and providers apply limits differently. Preserve that information and follow the applicable guidance rather than repeatedly submitting work. This explains a possible delay; it does not prove the status of every pending message. See the HTTP 429 reference.
If there is no explanatory evidence, say so. Absence of visible errors is not proof that processing is healthy, and lack of operator visibility is not proof that processing has stopped.
5. Check message relevance before waiting longer
A technically recoverable campaign may no longer deserve release. Separate the processing question from the business question: if the pending messages went out now, would they still be accurate, wanted, and actionable?
Review time-sensitive claims, availability, event timing, links, and instructions. Ask the campaign owner to define a relevance cutoff where one is needed. That cutoff is a business decision, not a queue timeout. In the fictional worksheet, noon is the owner's latest acceptable release time, not a platform requirement.
Also recheck recipient eligibility before recovery. Honour opt-outs and confirm the appropriate permission basis. For WhatsApp Platform messaging, approved templates are used to initiate conversations and outside the 24-hour customer-service window, which opens or resets with a user message. Template approval does not guarantee delivery. Check the current WhatsApp Business Messaging Policy for precise requirements.
If the message is stale, do not treat clearing the queue as success. Seek an authorised hold or cancellation where available, establish what remains in flight, and prepare any revised communication as a separate decision. Editing a message elsewhere should not be assumed to change work already submitted.
6. Escalate with a bounded question and usable evidence
A useful escalation asks someone to resolve an uncertainty. Campaign stuck is less actionable than: the same pending cohort has not advanced across fresh observations; can you establish whether it is waiting, blocked, or already handed off?
Include the campaign and workspace identifiers, observation timezone, submission time, successive counts, oldest pending reference, latest processing event, relevant errors, and the checks already performed. State whether new work entered the scope. Share only the recipient information needed for the investigation, through the appropriate authorised channel.
Distinguish what is known from what is inferred:
- Known: The available records show no new processing events for the remaining cohort between the recorded observations.
- Inferred: The remaining campaign work may be stalled.
- Unknown: Whether an unobserved downstream handoff occurred.
- Decision needed: Continue waiting, hold remaining work, or authorise a narrowly defined recovery.
Assign an owner and a next review time even after escalation. Otherwise, responsibility can disappear between the operator, campaign owner, and support team while the message becomes less relevant.
7. Authorise recovery only for work with a known disposition
The most damaging shortcut is resubmitting the entire campaign because the first submission looks inactive. Some messages may already have progressed even if the campaign summary has not caught up.
Before recovery, identify which items are confirmed complete, explicitly failed, still pending, or unknown. Establish what an available retry or resubmission action actually affects. Do not assume it excludes previously processed recipients or stops the original submission.
A recovery authorisation should name the affected cohort, permitted action, responsible operator, relevance cutoff, and stop condition. Check the current workspace and plan for available controls; do not assume pause, cancellation, detailed exports, or selective recovery exist.
Where supported and explicitly authorised, a limited recovery can reduce the uncertainty introduced by a large restart. Its trade-off is additional reconciliation work. A successful limited attempt does not by itself prove that every remaining item is safe to repeat.
If an attempted recovery produces contradictory counts, unexpected processing, or ambiguous outcomes, stop further recovery actions and preserve the new evidence. Reconcile original and recovery work together. Repeated intervention without reconciliation makes it harder to distinguish delayed originals from new attempts.
8. End each check with a measurable next action
Every observation should produce a short decision record: current state, supporting evidence, action, owner, and next check. Use operationally honest labels such as progressing, suspected stalled, visibility unresolved, or no longer relevant rather than forcing every case into healthy or failed.
For the fictional worksheet, the next action is specific: at 10:00 UTC, the assigned operator will compare the same 480-item cohort with the 09:40 evidence and add any support finding. If movement resumes, they will verify what advanced and whether the message remains useful. If disposition remains unknown, they will keep related submissions paused rather than authorise a blind restart.
Close the incident only when the remaining work has a recorded disposition, any recovery has been reconciled, and the campaign owner knows what was delivered, failed, cancelled, or remains unknown. A smaller pending count is evidence to interpret—not, on its own, a reason to declare recovery complete.
Put the next conversation to work.
Explore Alactic Communications for WhatsApp conversations, campaigns and AI-assisted workflows. Choose the product and plan that fit your team.
Explore Communications