This is a tricky one.

I have a QSEoW 2020-Nov-patch1 environment from which I am getting strange timeout errors.
A couple of different virtual proxies: Azure SSO via SAML, and standard Windows auth from the login page provided by Sense.

Session length is 30 minutes and an idle timeout will result in the user being sent to the login screen. All good.

Most times a click on the "Login" button will result in the username/password (for Windows login endpoint) or the Azure login (for Azure SSO) flow being shown.

Sometimes users get this though:

I've been unable to pinpoint what's causing it and cannot reproduce it.
Those 400 errors come in various shapes and forms (for example:  if a user is arriving with a URL that's not whitelisted in the virtual proxy settings that will give a 400 error), but I've never seen one with "Bad Request" twice like this.

Also, I can't find anything in the logs. Zero.

Usually these kind of errors result in proxy log entries, not this time though.
I've searched all logs from all services, no luck.
I might of course have missed something... Hopefully I've missed something.

So: The issue happens often enough to be an annoyance and it's definitely a loose end.
And loose ends aren't any good.

Any ideas, anyone?

-----------
Edit:

Some additional context. There is a Traefik 2.3 reverse proxy in front of the Qlik Sense server, meaning that traffic to Sense pass through Traefik. The root cause of the issue may be in Sense, in Traefik, in Azure - or any combination thereof.

BUT: Given that the 400 error obviously is served by Sense, I figure there should an error entry somewhere in some log. In the best of worlds it would give a hint to what's going on.