Two problems: The definition of "uptime", the host SLAs only covers network uptime and environment uptime, but clients consider "uptime" to mean application le

Two problems: The definition of "uptime", the host SLAs only covers network uptime and environment uptime, but clients consider "uptime" to mean application level uptime, which includes downtimes for server maintenance, steady-state backups, backup restores, deploying new releases, etc ... anything other than service is not 100% fully functional = downtime in their minds. Also for costs, "reasonably fast failover to another similar host" implies live redundant equipment at another host which is double the hosting costs, that's a big pill to swallow, so big that most orgs would rather suffer the downtime when they see the real cost of full redundancy.