Overview page shows 0 requests / no traffic data on some projects (silent AbortError in root layout)

[alexnogueirasilva]

Summary

On a self-hosted instance (v0.5.0), the Project Overview page's summary cards ("Server Requests", "Unique IPs", "Avg Response", "Bandwidth Out") and the "Traffic" chart show 0 / N/A / "No traffic data yet" on some projects, while other projects on the same instance/session render correctly with real numbers.

This is not a data problem — the underlying analytics endpoint returns correct, non-zero data for the affected projects. It's a client-side rendering issue.

Reproduction

Across 5 projects on the same instance/org/session:

• mesh-api, eyris → Overview page renders real traffic numbers correctly.
• 3 other projects → Overview page consistently shows 0/N/A/"No traffic data yet", on every reload.

For an affected project, calling the exact same endpoint the page itself calls, from the page's own JS context (same session cookies, same fetch call the frontend issues):

await fetch('/api/proxy/api/analytics/overview?projectId=<id>&domain=<domain>', {credentials:'include'})
  .then(r => r.json())
  .then(j => j.data.summary)
// => { totalRequests: 3057, uniqueVisitors: 3010, avgResponseTimeMs: 63970, ... }

...returns correct real data. So the API layer and the exact request the frontend makes both work fine. The page just doesn't render it.

Console error (only fires on affected projects)

On every load of an affected project's Overview page, and never on a working one, the console throws:

AbortError: signal is aborted without reason
    at app/layout-<hash>.js:31:3830

This fires from the root layout.js chunk (shared across all routes), suggesting a shared data-fetching hook's AbortController is being aborted before the request completes — likely a race condition tied to some per-project state, since it's consistent per-project (always fires for the same broken projects, never for the working ones) rather than random.

Impact

Users lose visibility into real traffic/usage for affected projects directly in the dashboard, even though the data is being collected and is available via the API. Currently working around it by querying /api/analytics/overview directly / using external monitoring instead of the dashboard widget.

Environment

• Self-hosted, release v0.5.0
• Reproduces consistently across multiple browser sessions/reloads on the same affected projects

[Farahat612]

Adding a data point that may help both this issue and #421: the reason the overview request is slow enough to hit the client timeout on some projects is that the tab fires a domain-LESS analytics request one render before the selected domain lands (child effects run before the parent's). The API answers a domain-less request by fanning out over every tracked domain, one bounded edge round-trip each. On multi-domain projects that discarded request is the expensive one and the real request queues behind it into the abort. So the timeout raise in #421 treats a real symptom, but there is also an avoidable 2x request (one of them N times more expensive) underneath. I have a branch that derives the selected domain on first render so the wasteful request is never issued, plus a bounded retry for transient aborts in the shared hook; happy to open it as a complement to #421 (it does not touch the error-card work) if useful.