Bingo, and it shows a fundamental design flaw in Wildfire.
From PAN's documentation on their latest OS: https://docs.paloaltonetworks.com/wildfire/9-0/wildfire-admin/wildfire-overview/about-wildfire.html (emphasis mine):
The WildFire Analysis Environment identifies previously unknown malware and generates signatures that Palo Alto Networks firewalls can use to then detect and block the malware. When a Palo Alto Networks firewall detects an unknown sample (a file or a link included in an email), the firewall can automatically forward the sample for WildFire analysis. Based on the properties, behaviors, and activities the sample displays when analyzed and executed in the WildFire sandbox, WildFire determines the sample to be benign, grayware, phishing, or malicious. WildFire then generates signatures to recognize the newly-discovered malware, and makes the latest signatures globally available every five minutes. All Palo Alto Networks firewalls can then compare incoming samples against these signatures to automatically block the malware first detected by a single firewall.
Sure, we could create a signature to block their non-malicious test file.
However, it'd be an Anti-Virus signature, exactly the same way PAN blocks files detected as malicious by Wildfire.
If you subject our demo doc to Threat Emulation, a report will be generated.
However, you will see blocks by URL Filtering and/or Anti-Virus as well 🙂
It mentions "Reputation" as well as an "Exploited Macro" as the reasons for blocking.
It's not actually a malicious file, of course.