Why are relays so frequently driven by optocouplers?
First, a possibly more permanent link to this product is here. And the schematic is here. (Edit 7/29/2015: Ironically my two links are now broken and OP's Amazon link is still useful)
Two reasons it makes sense to use optoisolators here:
The controlling device might be very far away so that it doesn't share a common ground reference with the relay board (except as connected through a long cable). Using the optoisolator means the control signal is used purely as a differential signal between Vcc and the control signal, both sourced from the controller circuit; ground potential differences won't affect the operation.
The relay coil voltage is not necessarily the same as the controller's Vcc. It could even be generated by an off-line (unisolated) supply. The optoisolator then provides isolation between the potentially unisolated
JD-VCCsupply and the controller circuits.
Probably a number or reasons, but the most important being that it will prevent transient voltage from damaging the driving transistor. And depending on the application, it will help prevent AC noise from interfering in the rest of the circuit.
You bring up some good points, however optocouplers are commonly used to isolate components from potentially dangerous outside sources. They are cheap and simple to implement. And they can potentially offer more protection than a diode. And of course, as you pointed out:
A number of these boards don't seem to be designed brilliantly (no regard to clearance or creepage), so even if the optocoupler is simply to provide two layers of isolation, the board fails at this.
I suspect a big part of the reason has to do with the idea that if there are two isolation barriers, there will continue to be an isolation barrier even if one is accidentally or intentionally bridged. When working with circuits, especially if one is a klutz, one may sometimes briefly short things which really shouldn't be shorted (e.g. because a scope ground clip decides to come undone and flail itself across the board). Adding an extra layer of isolation reduces the likelihood that such an accident will cause significant damage to anything. Most mass-produced products will never be on anyone's workbench, much less a workbench belonging to a klutz, but many home-brew products will spend a great deal of time on such workbenches. Further, home-brew boards are often made without solder mask, greatly increasing the likelihood that of stray ground clip or probe making an unwanted contact.
In addition to providing protection against accidental bridging, if there are two full isolation barriers it may be possible (if one is careful) to bridge one while doing diagnostics involving the other while maintaining an isolation barrier between the two main parts of the system. For example, if one wants to determine the amount of time that elapses between the processor setting an output and a solenoid receiving power, one could start by confirming the relay-coil ground and contact-side ground were isolated, bridging the relay ground and CPU ground, and measuring the time between the CPU output and the relay coil. One could then isolate the relay-coil ground and CPU ground and--after double-checking that they really were isolated, bridge the relay-coil ground and contact-side ground and measure timings between the coil and the things they control. Performing such measurements in a system with only single isolation would probably require having a scope with two probes that were isolated from each other. Such rigs exist, but they're generally expensive.