Yeah I read that about an hour after I posted my post :-) Both 6.7 and 7.0 say 30%.
Them 30% is a bit "conservative" eh? Customers always complain about needing to buy so much % more storage than the usable capacity that they need and vSAN throws salt in that would by requiring so much "overhead" (as that's how customers tend to interpret it).
On a normal cluster, this makes sense as there needs to be space on other nodes for rebuilds in case of "node-death". On 2-Nodes however, there is no place to go so those 30% and "What if the most consumed host fails" Alert are rather pointless. I'd say get rid of the latter alert on 2-node and up the free space threshold to 85% or so...