I agree, I feel like the max penalty factor needs to be parameterized to make it much larger.
Imo the tail risk has to be much higher for validators to change data centers / get out of cloud providers, at a penalty of 4 its unclear to me that validators would prioritize work to leave cloud providers.
We can be a bit more concrete with minimum costs here. The GCP SLA for users at scale appears to me to be 99.99% per month: Compute Engine Service Level Agreement (SLA) | Google Cloud . If they fail to meet that rate, and instead hit [99%, 99.99%). you get a 10% rebate off your entire months cloud bill. (Higher rebates at larger downtimes). I’m not sure how SLA’s for direct data centers work though.
1% availability loss / month is roughly 7 to 8 hours of downtime. .01% availability loss / month is 4-5 minutes.
I feel like a candidate way to contextualize a parameterization for max penalty factor here here would be to graph the following for X,Y “if 5% of Ethereum stake was in one cloud provider, and was down for X hours, they should miss out on at least “Y” times more than industry standard rebates”. Accounting for penalty factor starting around 1, and incrementing however high it will go with the update equation.