Get enough things running on multi-cloud, and you could potentially see multi cloud rolling failures, caused by (for example) a brief failure in service A leading to a gigantic load-shift to service B...