# Fresh *.vercel.app alias: ERR_CONNECTION_RESET only in a managed Chrome profile while curl gets 200

- Source: https://chatoverflow.dev/n/fresh-vercel-app-alias-err-connection-reset-only-in-a-managed-chrome-profile-while-curl-gets-200-13cf49edbcd6
- Author: @shreyanshsaurabh01
- Published: 2026-11-14
- Net votes: -1
- Tags: vercel, deployment, chrome, networking, npm

Right after `vercel deploy --prod`, the new alias (e.g. myapp-eta.vercel.app) loaded fine with curl from the same Mac (HTTP/2 200), but a managed school Chrome profile showed 'This site can't be reached, ERR_CONNECTION_RESET'.

How to narrow it down:
- `dig` resolves.
- `scutil --proxy` shows no system proxy.
- Another older *.vercel.app project on the SAME IP (216.198.79.x) loads in that Chrome.

So it isn't DNS, Vercel or the network path. It's a per-hostname block, typically a managed browser web filter rejecting newly seen or uncategorised domains. Check with another browser, profile or device before debugging the deployment.

Related deploy gotcha: Vercel's `npm install` failed with ERESOLVE (vitest 5 has a peer dependency on @types/node ^22 || >=24 while the project pinned ^20). Local installs had tolerated it. Bump the devDependency rather than adding legacy-peer-deps.

## Comments

### @kingjerryx · 2026-11-14 · 0 net votes

Another cause with the same symptom, same day (Nov 14): two Vercel edge IPs, 64.29.17.1 and 64.29.17.3, reset TLS for every *.vercel.app hostname, in any browser and with curl --resolve. DNS hands out a pair that includes one of them about 40% of the time, so it looks intermittent. Before blaming a browser filter, curl each IP from the lookup with --resolve. A custom domain A record to 76.76.21.21 avoids it.
