I tend to work with my clients for the lifetime of their projects. Often, the systems I develop become vital business components and, for ISO/quality compliance or simply for reassurance, they require general terms of service and details of provisions for disaster recovery, including the possibility of something unfortunate happening to [me] the developer.
Put simply, unless otherwise agreed, clients own their content and data, and the source code remains the property of the developer, with the software supplied as a service, with an implied or formal licence.
Some hosting companies offer a service level agreement that compensates customers based on the amount of any down time. Hardware can fail causing routing issues or server crashes, and while outages are rare, they are impossible to rule out 100%.
Based on a hosting fee of £100, the unscheduled downtime of Typomania services over the past few years would equate to an average compensation amount of less than 5 pence.
All files and data are backed-up during the night to 2 separate servers, the first being a live, redundant server on standby to take over in the event of an unrecoverable hardware failure, and the second being a failsafe copy.
In the event of an outage, we will first try to establish where the problem is, as most outages tend to be routing issues. If the problem is identified as a server fault, remedial action will be attempted. If unsuccessful, domain names can be pointed to the primary back-up server which will take over as things were at the end of the previous day.
For business data systems, the back-up frequency can be increased to hourly, to reduce the maximum potential data loss. In special circumstances transactional mirroring can be enabled to that database calls are transmitted to a back-up server asynchronously, a few moments after the primary action.