So in late November, a panel of 26 of us ended up forecasting a 76% chance that Twitter would have an outage by Jan 30, which happened. Wrote about it here: https://magoo.github.io/risk-measurement/blog/forecasting-a-twitter-outage/
Condition #1 easily passed - several newspapers of record called it a widespread outage.
Condition #2 passes, though, DownDetector being the "measurement" that the newspapers cited just barely passes the rules as written.
Condition #3 was easy - could not read or write tweets.
The rules were written to capture a real gnarly outage, and this one sorta squeaked by right over the bar.
In the future, we might be able to use panels like this for more objective measurements of downtime: https://deadbird.singlepane.io/d/hI9vrUO4k/home?orgId=4&refresh=30s
Thanks to all the 🍕 panelists who participated!