The flawed presumption is that all hardware RAID recovery algorithms are aggressive and/or can't be parameterized. If Drobo is being shipped with those drives then most likely Drobo has adapted to the slower error recovery mode of the more mainstream drives they (and except their customers to put) into their systems.
The higher end hardware RAID vendors don't. They expect and have optimized to different set of drives with different characteristics. The root cause of the failure here the mismatch between the enclosure vendor's design expectation and the customer sticking in drives not on the certified(or designed for) list.
The core issue is how many time the drive will try internally to correct for an error. Most mainstream drives are used where there is no alternative source for the data ( other than perhaps a slow back-up). They will try multiple times to work around a read/write error before giving up. Enterprie drive in RAID mirror/parity context the alternative version of the data is somewhere else. So if those drives quickly punt, the RAID controller can just quickly look somewhere else and come back to this drive to see if it continues to have problems. If the drive doesn't respond quickly it will assume the entire drive is bad and basically kick it out.
If in a parity/mirror set up can probably survive one drive being prematurely kicked out but if kick out two then the whole RAID set goes bad ( or at least dark ).
----------
For RAID yes ( depending upon workload needs). For JBOD (or software RAID 0/1/10 layered on top ) not so much. In the second case there are $/performance issues. It works, but probably paying paying alot more than have to.
The "49'er gold rush" to the Promise T2 solutions is questionable. There is probably going to be a limited window where it is the only shipped TB v2 solution. "Only" is different from "VERY GOOD".
