There is a fixed catalogue of things KonnectHub will tell you about. Each entry decides its own urgency, which channels it uses unless you say otherwise, and who is eligible to receive it.
| Notification | Raised when | Urgency | Default channels |
|---|---|---|---|
| Deployment is failing | A deployment fails repeatedly, or its success rate collapses | Critical | In-app, email |
| Deployment recovered | A failing deployment is running cleanly again | Info | In-app |
| Deliveries are being held | A switched-off deployment starts holding work instead of running it | Warning | In-app, email |
| Connection broken | A connection can no longer authenticate | Critical | In-app, email |
| Connection issue reported | A provider refused part of a connection — a topic, a scope, a webhook | Warning | In-app, digest email |
| Connection recovered | A broken connection works again | Info | In-app |
| Invitation accepted | Someone accepted an invitation and joined | Info | In-app |
| Your role changed | Your role in a workspace was changed by somebody else | Info | In-app |
| Removed from a workspace | You were removed from a workspace | Warning | In-app, email |
| Access token expiring | A personal access token is close to expiring | Warning | In-app, email |
| Message from a flow | A flow raised one itself | Info or warning | In-app |
| A delivery channel was disabled | Your webhook failed too often and was switched off | Warning | In-app, email |
| Test notification | You pressed "Send a test" | Info | In-app, email, webhook |
Critical is not a louder word for important: it is the only urgency that ignores quiet hours. Exactly two kinds carry it, and no flow can raise one.
Health is judged per deployment, over a rolling window of its recent finished runs. Cancelled runs are not failures, held deliveries are not runs at all, and draft runs are excluded completely.
A deployment is failing when it has failed several times in a row, or when enough of its recent runs have failed to put its success rate on the floor. It is degraded when it is losing a smaller share — visible as a badge on the deployment, and silent by design. Anything else is healthy.
Only crossing into and out of failing is worth a notification:
So a deployment failing every minute for a day produces two notifications: one when it broke, one when it came back.
A connection that stops authenticating raises connection broken — critical, because every flow using it will fail from that moment.
Providers also refuse things individually: one webhook topic, one scope, one renewal. Those are recorded as issues on the connection and raise connection issue reported the first time each distinct problem appears, not every time a background reconcile notices it again. Issues at warning level are shown on the connection page and raise nothing.
Both clear themselves when the connection works again, and a recovery notification follows.
Switching a deployment off does not throw its work away — each delivery is held. The first hold raises one alert; further holds keep its count and the age of the oldest delivery current without nagging you again, and it re-marks itself unread about once a day while the backlog is still growing.
It clears when the backlog empties — a release or a discard — and deliberately not when the deployment is switched back on, because turning it on does not dispatch the backlog. The decision this alert exists to ask for is still waiting.
Invitations accepted, role changes and removals go to the people they concern; the person who made the change is never notified about their own action. Removal from a workspace is addressed to your account rather than to the workspace you just lost — otherwise the one person who needs it could not see it.
Personal access tokens warn their owner shortly before they expire, once per token.
A flow can raise a notification itself, for something only your code knows — a supplier feed that arrived empty, a reconciliation that found a mismatch. They all arrive as message from a flow, so one switch quiets every flow that talks too much. See Notifying from a flow.