Thanks for making a pull request to jupyterlab!
To try out this branch on binder, follow this link:

The console test failures appear relevant.

The remaining test failures are not related.

Thanks!

For browser with non-default drive (like for RTC), the kernels should start in the file folder and the terminals and consoles should open in the current directory.

What does "file folder" means here? What should then be the expected behavior in the case when multiple drives with possibilly the same file names (example.ipynb in both the default filebrowser and another drive) are used simultaneously?

With 3.6.x it was possible to get some information about this since the drive name was kept as the session context prefix, and visible in the tooltip:

But with JupyterLab 4 it's not the case anymore:

Which makes it difficult to know about which kernels are being used, especially when there are files with the same name in two different drives:

Raising this question because it looks like this change (released in JupyterLab 4.0) affected JupyterLite downstream: jupyterlab-contrib/jupyterlab-filesystem-access#55 (comment)

The drive name used to be passed along as a prefix (for example FileSystemAccess:) which means the server was aware of the drive currently in use, which could then be passed to the kernel on startup. This allowed for accessing files from other drives from within a JupyterLite kernel in a seamless way.

Looking more into it, JupyterLab does not even seem to allow this anymore, since the information about the drive being used in never sent to the server, now that the session context path defaults to the local path (without the drive name).

Potentially something to look into as part of #15143, to allow for such uses case like in JupyterLite.

Looks like #13866 might also be related.

What does "file folder" means here? What should then be the expected behavior in the case when multiple drives with possibilly the same file names (example.ipynb in both the default filebrowser and another drive) are used simultaneously?

I have similar questions as @jtpio.

Overall, my question is what is the strategy/logic of the drives (registration, usage, naming convention, URI...) across JupyterLab and its extensions.

This is a broader question than the issue fixed by this PR, and this may be already defined somewhere (?), just asking...

This is a broader question than the issue fixed by this PR, and this may be already defined somewhere (?), just asking...

Yes very likely a broader question that would involve more design work on the drive feature. I opened #15289 to discuss this more in an issue.