Customer Telemetry and VPS Console Reliability
Live VPS and bare-metal telemetry now updates correctly, and cold browser sessions can load the complete Proxmox console without partial asset failures.
We repaired two customer-management reliability problems affecting live resource readings and browser console sessions.
The CPU, memory, and bare-metal hardware widgets were successfully rendered, but their browser polling code was not authorized by the site's enforced Content Security Policy. The scripts now receive the same response-specific nonce as the policy, allowing them to request authorization-scoped telemetry without weakening CSP.
The Proxmox console issue occurred before a console bridge was created. A normal cold browser cache loads noVNC as a graph of many small JavaScript modules. Those assets had inherited a low-volume form-submission rate bucket, causing later dependencies to receive temporary 503 responses. The console location now relies on the site's normal request and connection controls instead of the form bucket.
We validated the result with live customer-scoped Proxmox and iDRAC telemetry calls, all 41 noVNC modules through the public edge, and an end-to-end VM console WebSocket handshake through the WAF, Nginx, console gateway, and Proxmox. Existing iDRAC console behavior was preserved.