Fault tolerance exists on several levels
Second is the build node. The build controller will periodically poll build nodes to test their connectivity. If one node is lost, it is taken out of the pool. Any pending builds on such a system can be re-queued by the developer or automatically by the build controller
Source control and Artifact DB - these usually exist on systems with redundant and fast disk arrays that maintain daily backups. These systems can usually handle single hardware failures and continue to operate.
To the developer, the only thing that exists is the source control system and the build controller.
The build controller manages a queue of “builds” that a developer monitors.