Hello Vercel team,
The Vercel Support assistant directed me here because it could not open a human support case for this issue on Hobby. I would appreciate help understanding the CPU usage and the supported no-cost steps to restore service after remediation.
Dashboard snapshot reviewed on November 19 (Oct 20-Nov 19, 2026; rounded values):
- Fluid Active CPU: 13h 10m across the team. One Nuxt application accounts for about 10h 19m (78.4%).
- Temporary 3x limits increase until Nov 20. The normal CPU allowance is 4 hours; 3x would be 12 hours, so the reported total exceeds both.
- Team paused since Nov 18, with Dec 18 displayed as the pause end date.
- Fast Data Transfer: 5.41 GB / 100 GB normal allowance.
- Fast Origin Transfer: 2.86 GB / 10 GB.
- Function Invocations: 405K / 1M.
- Provisioned Memory: 179.7 GB-hours / 360.
- CDN Requests: 602K / 1M.
I understand these resources have independent limits and low bandwidth does not imply low CPU. I am not claiming the meter is wrong; I would like to understand when and why Active CPU reached the cap, and the exact threshold that triggered the pause.
Investigation and fixes:
- Nuxt server routes are grouped under /__fallback. The available recent function metrics did not let me attribute the historical monthly CPU total to individual handlers.
- Code review found repeated cleanup SQL during legacy session reads, repeated schema initialization, repeated Intl.DateTimeFormat construction, and background reminder/Google Tasks schedulers starting on server instances. These are possible contributors, not proven causes of the historical increase.
- Local fixes reduce legacy session reads from four database queries to one, with zero queries when there is no session token; reuse collection schema initialization; and reuse date formatters.
- 24 targeted tests and a local production build passed. At the November 19 review, these fixes had not been deployed or measured live. Earlier public-page prerendering and background-refresh optimizations also appeared newer than the inspected production deployment.
- I understand that less database waiting does not directly translate to the same reduction in Active CPU, and local microbenchmarks cannot establish the future monthly total.
Could a Vercel staff member please help with:
- Confirming the actual pause threshold and timestamp, and when the temporary 3x allowance started and applied.
- Identifying any available daily CPU/deployment breakdown, unusual traffic, cold-start activity or background work that explains the increase, without requiring a paid observability upgrade.
- Confirming whether remediation can be reviewed for a no-cost unpause, or the correct free restoration procedure and how to deploy and verify the fixes afterward.
- Clarifying whether the displayed Dec 18 end date requires a manual resume action. The documentation says paused teams do not automatically resume.
I want to stay on the free Hobby plan and am not requesting or authorizing a paid upgrade, trial, add-on or charge. Please let me know if account-specific information is needed through an official private channel; I have omitted team IDs, project names and domains from this public post.
Attached: screenshots of the CPU history grouped by runtime and the networking overview, captured November 19. Account details and project names are excluded.
Thank you.