That is exactly the case and an important point, Jack. All RAID arrays have their own coding, and drives that are a part of that array must be properly formatted for the array. This can even be true for RAID 1 (mirror) arrays, meaning that if one drive in the array fails, the other drive will rebuild the array with the replacement. It can be a false security thinking that if one failed, you still have all your data readable from the other outside of the RAID array configuration. Only way to ensure that is to clone one drive to another as pure standalone drives.
The performance on the Drobo should be pretty decent, once it is allowed to format the drives, and get itself set up properly. This will take time, but then it takes time on any RAID array, so folks should not be surprised. (The videos are a bit misleading when drives are popped in and out. That part is possible, but the full performance capabilities will not be there until the system gets completely stabilized.)
Jack, using this device for efficient-redundant data storage would be ideal, but I also think it could be used in normal storage workflow once drives are in place, formatted, and the system is stabilized. The FW800 capability is more for date throughput, and will be "gated" somewhat by the internal processing. The USB2 channel will be a lot slower, as USB goes through the CPU. In this device, they are essentially installing a small CPU for that traffic, so the hit within the device will not be bad, but USB2 back on the main computer will be a different issue, and will be affected by the CPU usage of that computer. (This is one of the nicer things about FW....it is more or less able to handle data transfers without involving the main CPU of the host, and that is why it can have higher sustained throughputs than USB2.)
So folks do need to understand that this device has some great capabilities, and that it is also not going to be as fast as hardware RAID arrays, and that drives placed into its system will take time to build and organize. If you can live with those parameters, and let it set itself up properly, it should work very nicely and provide some level of confidence for data redundancy.....to a point.
LJ