Is replacing a failed hard disk as simple as turning off a node, swapping in a new disk and restarting the node?

Yes. Depending upon the customer, they may wish to let more than one disk fail before replacing the bad disks. For example, if the nodes are in a remote data center and access is cumbersome, a customer may wish to let a cluster lose multiple disks before sending out a repair person.

If the chassis is capable of hot swap Hot Swapping and Plugging Disks (single disk RAID 0 configurations, for example, are not capable of hot swap), then you do not even need to power down the chassis before swapping out the disk.