EBS is not a monolithic service, but the volumes are not individual disks either. Each volume is served from two (mirrored) servers. These servers are pretty independent. The outage apparently was caused by many of them being isolated on the network due to the flooding and independently deciding to create a new replica. EBS is still a great service and I wouldn't expect a mass exodus. It has always had its problems with inconsistent performance and interference of snapshots on performance. What I do expect to happen is users re-evaluating whether they need EBS or can do with the instance stores after all. For example, we have used EBS as store for a number of cassandra NoSQL stores and are re-evaluating whether that's the best choice. It's not an obvious trade-off.