WPT Inconsistent CDN use test results

I am having a hard time understanding some conflicting results I am getting from a test ran against the same page from 2 different locations, specifically in regards to the effective use of CDN.

The page, when tested from Salt Lake City, shows as Effective use of CDN.
https://www.webpagetest.org/result/210625_AiDcB5_9c41d14b21aa6e9709da9295fe142e40/

The same page, when tested from California. does not show us using the CDN effectively.
https://www.webpagetest.org/result/210625_BiDc0S_dc20186c38c424e3531315500202011c/

How can the same page’s use of a CDN results show different values based off of where it was tested?

I have validated that there is indeed a CDN (Azure CDN) being used for all of the static locally hosted items.

I looked at the response headers as well. The only difference I can see is that the test that does not show the page as using a CDN effectively is missing a header.
x-azure-ref-originshield: 0MYvUYAAAAAD5BO5SHZ6FR6uwC3+iOepeU0pDRURHRTA1MDkAMjg4NGMzZDktZmExOS00ZTliLTlhODQtZTZkMWVmNWIzZWE1

The rest of the response headers are identical between the 2 tests.

The x-azure-ref-originshield appears to be an Azure specific header that means the item was served up from the Azure origin shield layer.

Am I wrong in thinking that there is something in the network layout of the test machine that is somehow filtering out the header? Is the missing header the reason the test is not recognizing the Azure CDN?

My guess is something about the IP resolution from the 2 different regions cause it not to detect that it was using Azure’s CDN.

It looks like Azure currently only relies on CNAME matching for detection: wptagent/optimization_checks.py at master · WPO-Foundation/wptagent · GitHub

I’ll see if some of the headers can also be used for more reliable detection.