Suddenly seeing function timeouts with status code 504 when sending POST requests to storage provider supabase.com

[styxlab]

Steps to Reproduce

Note: You have to fill out a lot of forms until you'll get the the examples, so please look at the screenshots first. I can make a simplified example, but for now I'd just like to get some hints on where to look further. As the request looks fine on supabase provider side, and as it has worked flawlessly for quite some time, this very much smells like a Vercel infrastructure issue.

Getting 504 in vercel functions when doing POST requests to https://storage-api-canary-lb-eu-central-1-ext.storage.supabase.com. Logs in storage provider show that the request has been completed within 200ms, but vercel function does not complete with a 10 second limit and then errors with a 504 status code. Runtime is node. Also note that this function was working yesterday and I did not change anything in this function.
Vercel side:

Note: "before" has been executed, hanging in await statement

Note: The 504 is a secondary error, because of the 10 second timeout. Real question is: why is the await not returning, if the request is completed within 200ms as shown in the supabse.com logs.

Supabase side:

Code:

Note: "after" is never executed

[styxlab]

Solved it myself, by moving the supabase client instanciation to the request handler. Now, everything is working as expected again:

I am pretty sure, something changed in the Vercel infra. Would be great if you can dig out what infra changes are responsible and how they might affect other function invocations. I am using remix.run, btw.