No-Code SaaS Automation Platform · View 05 of 21 · People and journeys
What the map found
- The platform knew at phase 1 and the author learned at phase 2 — from a colleague. The gap between those two phases is the single worst failure in this product, and it is not a reliability failure: every component behaved correctly.
- Silence is the default failure mode of this platform, and silence has no natural end — which is why view 18 exists as a loop and view 17 carries a dedicated Silence row.
Where the architecture answers it
- A per-connection quiet-period expectation and its alarm (Section 4), so an unexpectedly silent source is detected rather than assumed healthy.
- Credential death classified as non-retryable with owner notification and automation pause after a grace period (Section 7).
- A repair inbox grouped by cause, and replay from the retained trigger event with the duplicate consequence stated before confirmation (Section 9).
- Rate-limited backlog release, so reconnecting does not fire a week of work at a provider in one burst.
Open
- Core Architecture Question 6 is unresolved here: whether reconnection should release the backlog automatically or re-ask the author. The map shows why it matters — by phase 5 the work may no longer be wanted.
