To help achieve the best possible price for you, we deploy our servers at a scale, pace, and expectation that helps us keep the costs low, to your benefit. However, this also means we defer maintenance on server failures until scheduled maintenance windows.
If the unfortunate situation occurs where a server fails, a replacement server is provided in your registration slot with the same OS and keys as the original was provisioned with. See also: Backups and Data Recovery.
Because of this, we have found the workloads that work best on these servers are for ephemeral scalable services which are spread across regions/pods, such as application farms.
The best parallel to this behavior are the lower-cost “Spot” instances at AWS, which have a lower price-point, but can be shut down as circumstances change.
As described prior, these servers are designed for distributed availability, so any single server is not 5-9. If you need 5-9 SLA, you can still reach that, through your software design. Setup your service crossing multiple zones, and you now have a higher availability, at a lower cost.
