Project: radiostations
Production URL: radiostations.vercel.app
My Hobby project was paused after exceeding the Fluid Active CPU limit. Usage jumped sharply starting ~June 10 and stayed elevated (~1hr/dayagainst the 4hr Hobby allowance) through the pause.
Likely root cause: around that date I shipped a live “now playing” polling feature (scheduled job via pg_cron, refreshing every ~20s) alongside a personalization engine computing preference scores per request. I’d already cut the cron frequency back once (from per-minute to daily) after noticing it was heavy, but the pause happened before that fully took effect.
Since then I’ve merged additional fixes to main:
- Homepage revalidate increased from 30s to 300s, cutting SSR re-render frequency 10x
- app/robots.ts added to block all crawlers from the *.vercel.app URL
These are committed to main (commit e771297) but have NOT deployed, since the project doesn’t build new commits while paused, so the optimizations are ready but stuck behind the pause itself.
Requesting a manual review/resume so these changes can actually take effect. I’m also willing to further reduce or pause the polling job entirely if that helps demonstrate the CPU usage is under control.
Happy to share more detail on the cron/polling setup if useful.