LinkedIn webhooks in production: duplicates, ordering, retries and timeouts
This guide covers the problems LinkedIn webhooks cause in production and how to handle each one. What LinkedIn sends and how it's signed is on the LinkedIn webhooks overview, every event type is on LinkedIn webhook events, and signature verification code is on Verify LinkedIn webhooks.
LinkedIn's webhooks are spread over separate products with separate access, and several delivery rules differ between them. Each section says which product a rule applies to, because LinkedIn documents most of them for one product only.
At a glance
| Problem | What to do | With Event Gateway |
|---|---|---|
| Timeouts | Answer the GET challenge within 3 seconds; return a 200 and queue POSTs | Answers the challenge itself and accepts deliveries at once |
| Duplicates | Use the key for your product, because only Community Management has notificationId | Drops exact repeats within a window of up to an hour |
| Ordering | Compare occurredAt or lastModifiedAt, and apply only what's newer | Logs every event and attempt, so you can see arrival order when debugging |
| Retries | Watch for blocked endpoints; pull what you missed where the product allows it | Keeps answering the challenge and stores deliveries while your service is down |
| Choosing events | Get access to the right product; track subscriptions for each member | Filters and routes events by type |
| Testing | Serve GET and POST on one HTTPS URL; don't use ngrok | Gives you a stable HTTPS URL and forwards to localhost |
| Other limits | Read every item in a batch; respect LinkedIn's data storage limits | Absorbs bursts in its queue and can strip fields with a transformation |
LinkedIn webhook timeouts and slow handlers
LinkedIn gives your endpoint 3 seconds to answer a GET challenge, and it documents no timeout for a POST delivery. Keep the challenge handler small, and return a 200 from POSTs before you do any work.
Why it happens: LinkedIn's Webhooks page asks for the challenge response with a 200 OK within 3 seconds, and it repeats the challenge about every 2 hours. A serverless function with a slow cold start, or a handler that fetches the secret from a secrets manager on each request, can miss that window and count towards the 3 failures that block the endpoint. For POSTs, LinkedIn asks only for a 2xx, so any delay limit you hit is undocumented.
Fix it in your handler:
- Handle the GET in its own short path: compute the HMAC of
challengeCodeand return JSON withContent-Type: application/json. - Load
clientSecretat startup rather than during the request, and avoid cold starts on this route. - Verify each POST and return a
200. The shared page accepts any 2xx, but the Talent pages treat anything other than200 OKas a failure, so a202isn't safe everywhere. - Push the payload onto a queue and process it in a worker. See why to process webhooks asynchronously.
- Plan the follow-up work too.
RESYNC_JOBSasks you to update the requested jobs within 30 minutes, and many Talent events need a second API call to fetch the details.
With Event Gateway: A source with the LinkedIn Source Type and your client secret answers the challenge on your behalf, so your service only handles notifications. Hookdeck accepts each delivery, then queues events and delivers them at a rate your service can handle, so your handler can take as long as it needs.
Duplicate LinkedIn webhooks
LinkedIn says a notification can be delivered more than once, but the right key depends on the product. Only Community Management payloads have a notificationId.
Why it happens: Retries are one cause, and LinkedIn's own subscription model is another. On Community Management, each subscriber to an organization gets an event with the same notificationId and a different subscriber, so the id alone drops real deliveries when one webhook serves several members. On Lead Sync, leadGenFormResponse is reused for the same member and form, so a member who registers, unregisters and registers again produces three real events with one URN. On Talent push events, each retry has a unique id, so that id can't identify a repeat.
Fix it in your handler:
- For Community Management, walk
notifications[]and key each item onnotificationIdandsubscribertogether. Don't test this with LinkedIn's sample payload, which repeats onenotificationIdacross three different notifications. - For Lead Sync, key on
leadGenFormResponseandoccurredAttogether, as LinkedIn recommends. - For Apply Connect, key on
jobApplicationId. - For
EXPORT_CANDIDATE_PROFILEand other push events, fetch the record with the id, then deduplicate on what comes back, as LinkedIn advises. - For
JOB_POSTING_STATUS, treat each delivery as the current status ofexternalJobPostingIdand overwrite what you hold. The sample has no delivery id. - Make every handler idempotent as a fallback.
With Event Gateway: Hookdeck deduplicates events within a window of up to an hour. Use exact deduplication, which compares the whole payload, so that notifications with the same id for different subscribers aren't treated as repeats. Community Management retries can run for 8 hours, which is longer than the window, so keep the checks above.
Out-of-order LinkedIn webhooks
No LinkedIn page promises a delivery order, so compare a timestamp in the payload and apply only what's newer.
Why it happens: A retry 5 minutes later can arrive after a newer event for the same record. It matters most where a later event undoes an earlier one: a Lead Sync DELETED for a response, a COMMENT_DELETE for a comment, or a job's status changing twice.
Fix it in your handler:
- For Lead Sync, store the latest
occurredAtfor eachleadGenFormResponse, and ignore an action older than the one you have.occurredAtis what separates a member's repeated actions. - For Community Management, do the same with
lastModifiedAton each notification. - Accept a
COMMENT_DELETEorCOMMENT_EDITfor a comment you haven't seen, and record it, rather than failing the delivery. - For
JOB_POSTING_STATUS, the sample has no update time. If an old status would cause harm, read the current one from the Check Job Posting Status API before you act.
With Event Gateway: Hookdeck logs every event and delivery attempt, so you can see the order things arrived in when you're debugging. Your handler still needs the rules above.
LinkedIn webhook retries, blocked endpoints and lost events
LinkedIn's retry behaviour differs by product, and a separate rule can stop delivery altogether: 3 failed re-validations in a row block the endpoint.
Why it happens: LinkedIn re-validates every webhook URL every 2 hours. After 3 failures in a row the endpoint is blocked and events stop, so a few hours of downtime or a bad deploy of the GET route can end delivery while your POST handler is fine. Retries then work differently in each product:
| Product | Retries | After retries |
|---|---|---|
| Community Management | Every 5 minutes for 8 hours | Pull missed notifications, kept for 60 days |
| Apply Connect | 6 re-sends after a non-200 response | The applicant gets a delivery failure email |
| Talent push events | 5 retries, 300 seconds apart | The notification is dropped |
| Job Posting callbacks | Retried, with no schedule published | Not documented |
| Lead Sync | Not documented | Read leads from the Lead Form Response API |
Apply Connect failures cost the most. If the first application fails to deliver, the job converts to off-site apply, and failed validation does the same. A SIGNATURE_MISMATCH response is treated as fatal and isn't retried, so a verification bug loses the application.
Fix it in your handler:
- Send LinkedIn's warning emails somewhere that gets read. It emails the app's developers about each failed validation and again when the endpoint is blocked, and the portal shows the endpoint as Blocked.
- Monitor the GET route as well as the POST route, and test it in every deploy.
- Once it's fixed, re-run validation from the Developer Portal. Parent apps with Apply Connect or Job Posting integrations can also find blocked child webhooks and revalidate them with the DeveloperWebhooks API.
- Back-fill after an outage. On Community Management, query
organizationalEntityNotificationsfrom the lastlastModifiedAtyou stored. On Lead Sync, query lead form responses withsubmittedAtTimeRange. - On Apply Connect, return the matching
errorCode, so LinkedIn knows whether to retry.
With Event Gateway: Hookdeck keeps answering the challenge and accepts deliveries while your service is down, so an outage on your side doesn't count towards a block. Events are stored and retried on a schedule you control, and you can replay failed ones once a fix is out. See dead-letter queues for webhook reliability.
Choosing LinkedIn webhook events and keeping subscriptions alive
Which events you can get depends on which LinkedIn product your app has access to, and on Community Management a subscription belongs to one member and can end without warning.
Why it happens: Webhooks are only available to apps with an approved use case, and each product registers URLs its own way:
| Product | Events | Access and registration |
|---|---|---|
| Lead Sync | LEAD_ACTION | r_; POST /rest/leadNotifications |
| Community Management | ORGANIZATION_ | rw_organization_admin; Webhooks tab, then PUT /rest/eventSubscriptions for each member |
| Apply Connect | EXPORT_JOB_APPLICATION | Partner integration; jobApplicationWebhookUrl in the job payload |
| Job Posting | JOB_POSTING_STATUS, RESYNC_JOBS | Partner integration; Webhooks tab, one webhook per event type for each parent app |
| Recruiter System Connect | EXPORT_CANDIDATE_PROFILE, EXPORT_, INTEGRATION_, DELETE_PROSPECT_NOTE | Partner integration; callback URLs in customer configuration, unique to each customer |
| CRM Connect | EXPORT_CANDIDATE_PROFILE | Partner integration; callback URLs in customer configuration, unique to each customer |
A Community Management subscription is keyed on your app, the member, the organization and the event type. It's removed if the member's grant expires or is revoked, or if they stop being an admin of the organization, and LinkedIn skips a notification if those checks fail at the time of the event. Its docs describe no event for this, so you find out from API errors or from notifications that stop.
Fix it in your handler:
- Branch on
type, then onleadActionor each notification'saction, sincetypealone doesn't tell you what happened. - Don't expect every social action from private posts. LinkedIn only triggers
LIKE,COMMENTandSHAREfor posts withPUBLICvisibility. - Store when each member last received a notification, and query their active subscriptions on a schedule.
- Treat a 401 from an Organization API as a lost grant, which needs a new authorization. Treat a 403 as lost admin access, which LinkedIn says means the member can't resubscribe.
- Query the notifications missed during the gap, resubscribe, then use
notificationIdto skip anything you already pulled. - On Lead Sync, LinkedIn recommends subscribing per owner unless you need a separate URL per form or entity. Multiple matching subscriptions each get a delivery.
With Event Gateway: Hookdeck filters and routes events by type or any other property, so each of your services gets only what it handles. You still manage LinkedIn's subscriptions yourself.
Testing and debugging LinkedIn webhooks
LinkedIn needs a public HTTPS URL that passes its GET challenge before it sends anything, and it doesn't support ngrok, so test the challenge and the POST as two separate things.
Why it happens: Registration fails with This URL did not pass the security challenge check unless your response is right, and a route that only accepts POST can't pass. Since March 16, 2026, Lead Sync also pushes new leads only to validated webhooks. LinkedIn says most failures come from a wrong HMAC of the challenge code or the wrong client secret.
Test it:
- Serve GET and POST on the same URL. Return
challengeCodeandchallengeResponseas JSON, with the HMAC in lowercase hex. - Use the
clientSecretof the right app. If the GET has anapplicationId, it names the app to use. - Deploy to a serverless function on a cloud provider, which is LinkedIn's recommendation, or use a tunnel other than ngrok, though LinkedIn's docs don't say which ones it accepts.
- Create test leads from Campaign Manager. They arrive with
testLeadset totrueon the lead form response, though LinkedIn can't generate them for InMail campaigns with lead forms. - For Community Management, like or comment on a public post on a test organization. Re-run validation from the Developer Portal after you change the GET route.
With Event Gateway: Hookdeck gives you a stable HTTPS URL, and the Hookdeck CLI forwards events to your local server, so you don't need a tunnel that LinkedIn accepts. Give the source the LinkedIn Source Type and your client secret first, so that it can answer the challenge. You can inspect events in the dashboard and replay them against localhost without triggering new ones.
Other LinkedIn webhook limits
Batches. Community Management notifications are batched by event type with a batch size of 10, so read every item in notifications[].
Subscription cap. Lead Sync allows up to 1,500 lead notification subscriptions. Delete unused ones before you reach it.
Data storage. LinkedIn's Marketing API data storage requirements limit how long you can keep data you get through those APIs. Members' social activity data can be stored for 48 hours. Organizations' social activity data can be stored for six weeks, or six months if the organization has authenticated into your application. Lead gen response data has no limit. The limits apply to storing and caching, so check your logs and queues as well as your database.
Personal data. An Apply Connect payload can hold an applicant's contact details and voluntary self-identification answers, so treat these bodies as sensitive in your logs.
IP allowlists. LinkedIn doesn't share its IP ranges for push events and recommends a domain allowlist.
Short-lived ids. Talent push events can carry an expiresAt, and an application evaluation's requestId expires in 24 hours, so fetch the details soon after the notification arrives.
With Event Gateway: Hookdeck absorbs bursts in its queue and delivers at a rate you set, and a transformation can strip personal data before an event leaves your stack. Event data stays in Hookdeck for your plan's retention period.
Keep a failed re-validation from blocking your LinkedIn endpoint.
Hookdeck Event Gateway answers LinkedIn's GET challenge once the source has your client secret, verifies the signature, queues events and retries to your service on your schedule.