Custom domain routes to a stale deployment despite valid Vercel configuration

Hello Vercel team,

I have a custom-domain routing issue that appears to be serving a stale internal deployment.

Project: rakestra-site (team: rakestra) Current production deployment: dpl_3A6AFzuHwJ9AwLcjKmPzncngEB2c Current deployment URL: rakestra-site-j6ol2k2am-rakestra.vercel.app Git commit: 275a7e3d5be26007ae7898253152ff9f29ef8392

In the Vercel dashboard, both domains show Valid Configuration and are connected to this project's Production environment:

rakestra.com redirects (308) to www.rakestra.com
www.rakestra.com connects to Production

However, requests to https://www.rakestra.com/api/public-research/journey are served by an old Next.js Pages Router 404. Its response includes buildId XqkGNHt_ic9Y67qzHsjew. That does not match the current production deployment.

I have already tried, without changing DNS records:

Saving the existing www.rakestra.com Production connection again.
Removing and immediately re-adding both rakestra.com and www.rakestra.com to the same rakestra-site project.

The dashboard still reports a valid configuration, but the public custom domain continues to serve the old build. The deployment-specific URL has the expected current route (GET returns 405 because the route accepts POST), so the code was built correctly.

Could someone please check/reset the internal alias or edge routing for www.rakestra.com and identify which deployment owns buildId XqkGNHt_ic9Y67qzHsjew?

Thank you.

Additional evidence: this is not limited to the custom domain. The project's automatic alias https://playzonesports-site.vercel.app also serves the same old Pages Router build (including buildId XqkGNHt_ic9Y67qzHsjew) for /api/public-research/journey.

Rakestra

A comunidade dos esportes de raquete. Partidas, grupos e torneios no mesmo lugar.

By contrast, the current deployment-specific URL rakestra-site-j6ol2k2am-rakestra.vercel.app recognizes that route (GET returns 405 because it accepts POST), which confirms that the current deployment artifact is correct.

This makes the issue look broader than a single custom-domain record: a stale project/team alias mapping may be serving the old deployment. Could you please inspect/reset the project alias routing as well as the www.rakestra.com edge mapping?