Context is missing from parent loader when child's validateSearch fails

[Dakkers]

Which project does this relate to?

Router

Describe the bug

When a parent route (in my case, the root route) has a beforeLoad parameter, the parent route's loader's context is missing expected attributes when a child route's validateSearch fails.

StackBlitz project

StackBlitz container

Your Example Website or App

https://vitejsvitepmq9xxrp-by02--5173--96435430.local-credentialless.webcontainer.io/

Steps to Reproduce the Bug or Issue

1. Go to the example site
2. Go to the dashboard link
3. Ensure you can see the navbar (this is working)
4. Remove the query parameters from the URL and press enter (to navigate directly to /dashboard with no query params)
5. You will not see the Navbar anymore 🐛

Expected behavior

I expected the parent route's loader to have the values of the context that were returned by the same route's beforeLoad but the attributes are missing.

Screenshots or Videos

Screen.Recording.2025-07-18.at.10.00.33.mov

Platform

• Router / Start Version: 1.128.3
• OS: MacOS
• Browser: Chrome
• Browser Version: 138.0.7204.157 (Official Build) (arm64)
• Bundler: Vite
• Bundler Version: 7.0.5

Additional context

Related issues, PRs, discussions:

• Empty context with parent / child beforeLoad, loader and validateSearch #3578
• Sometimes the child route doesn't inherit parent context #3430
• fix: if validate search rewrites search params, issue a server-side redirect #4439
• fix: only re-execute beforeLoad if previous run failed #4433
• beforeLoad of child route is called even when parent’s beforeLoad throws an error #4151
• beforeLoad of children running before beforeLoad of the parent has finished #1737

[Dakkers]

My super awesome coworker @danconnell found out that this behaviour doesn't occur in 1.121.14, but it does on the Stackblitz instance which is 1.128.3, so I will do a bit more work to figure out when it started happening again.

[Dakkers]

It seems that the bug started occurring again in version 1.123.2. (1.123.0 had no issues.)

I would assume it's because of this PR because it makes modifications to beforeLoad / loader logic: #4550

[schiller-manuel]

could you add a failing unit test via a PR?