I have a 1M page site with export const revalidate = 2592000. Every deployment completely invalidates my ISR cache, causing massive ISR writes + costs.

Evidence:

1. Dec 31 had 4 deploys → 2M (!!!) ISR writes (~$8) with only 20 actual site visits on the day. Jan 1 already has 500K writes in 1 hour. I dont even have 2M pages!

2. I tested a specific URL before/after deploy - etag changed from “dyw7c1hf012e6g” to “3ivi8uvt5k2e64” and went from HIT to MISS. The trigger is changing translation files in public/locales/*.json.

3. Vercel analytics show 98% cache miss rate (257K misses / 272K requests).

My config: export const revalidate = 2592000, data wrapped in unstable_cache with 30-day revalidate, firewall rules active for bots. The docs say ISR cache persists across deployments, but mine clears completely every time. Is this expected? How do I prevent unnecessary ISR writes? What am I missing? My build logs show 30d revalidation but my cache still clears every time. I dont have any cache clearing configs in my vercel project settings.

Im on the pro plan. I’ve been trying to get support on this since Dec 12. I know its a busy month but I urgently need some help around this because of the extremely large number of ISR writes happening daily ballooning up my on demand costs.

Team URL: https://vercel.com/irfanahmds-projects

BUILD LOG SHOWS 30 DAY REVALIDATE

LAST 24 HRS FROM OBSERVABILITY

CURL BEFORE DEPLOYMENT (CACHE HIT - 01 Jan 2026 03:22:36 GMT)