Cloud Concentration Risk: Multi-Cloud Is Still Not Solved
The Monzo Stand-In thread (49417298) surfaced a comment that landed hard: 'By using another US hyperscaler for your backup platform you're still unhedged against US governance failure/breakdown.' Monzo built a fallback system for AWS outages, and the response from the community was partly admiration and partly 'you've just replaced one single point of failure with a slightly different one.' The AWS fake billing bug thread (48945241), where users got alerts showing bills of $109 billion, added texture. AWS billing infrastructure itself failed, which is a different category of failure than a service going down.
This is a pattern that comes up on HN repeatedly but feels more urgent now. The Google Cloud suspension thread (32547912) and the Google Cloud outage thread (20077421) are older but structurally identical: a single provider makes a decision or has a failure that cascades into your production system with no recourse. The Monzo approach, building a 'stand-in' mode that degrades gracefully, is the right direction but it requires significant engineering investment.
The counter-argument in the threads is that true multi-cloud is operationally expensive enough that most companies rationally accept the concentration risk. That tension isn't resolved.
So what?
If your business is meaningfully dependent on a single cloud provider, you should be able to answer 'what happens if they suspend our account on a Saturday at 1am' with something better than 'we go down.' Even a read-only degraded mode or a second provider for your most critical services is worth the engineering cost. The Monzo Stand-In architecture is worth reading as a case study.