Usually there's a hidden config register where we can change "the number for 911", so for some portion of testing, we'll set that to "Chris's phone" and inundate a coworker with calls, but once we've been through a shakedown with that, it's time to coordinate with the local PSAP and identify a slow time when they can field some real test calls from us.
There are all sorts of safeguards here -- they want a regular phone number for one of the testers, so if the test phone malfunctions in any way (say, calls complete but audio doesn't open), they can reach us out-of-band. And it's always perfectly clear that they can cancel the activity at any time, if enough real 911 calls come in that they don't have spare operators for our testing.
It's been a long time since I was on the carrier side, but I remember doing tower turn-ups, we'd also place one test call per carrier-sector, to make sure the location info was transferred correctly. Similar coordination but it was largely a formality because it was usually just 3 or 6 calls, and very brief ones at that.
In the instant case, both the phone maker (Google) and the carrier (unsure) should've exercised this in their acceptance testing. Since it sounds like a Teams bug, it's understandable that they didn't catch it, but also unacceptable that Teams could interfere with the ability to dial 911. Since we've now learned that apps can interfere with that, this hugely balloons testing requirements until/unless they fix the OS so apps cannot interfere.