The Drobo uses RAID just like almost every other device out there, they just hide it behind some ugly hacks. Since you started with 3x2TB drives, and 1x1TB drive, the Drobo was striping together 2 RAID 5 sets. The first RAID 5 used 1TB from 4 disks, then the second RAID 5 used the remaining 1TB of space on the 3x2 disks.

When you yanked the 1TB and replaced it with the 2TB, the Drobo will keep the 2 RAID 5 setups around, expanding the second one to take advantage of the new space on the new disk. It's having to rebuild the first one, as 1TB of the 4TB is gone, so this involves lots of checksum calculations to piece together all the missing data. It then has to restripe the second RAID 5 to use 4 disks instead of three, so the entire second RAID set has to be recalculated for the new setup.

Here is some ASCII art to try to explain it better:
Code:
From this:
         ----------
         |   A1   | One RAID 5 partition (1 TB)
         ----------
-------------------
|   B1   |   A2   | Two RAID 5 partitions (2× 1TB)
-------------------
-------------------
|   B2   |   A3   | Two RAID 5 partitions (2× 1TB)
-------------------
-------------------
|   B3   |   A4   | Two RAID 5 partitions (2× 1TB)
-------------------

to this:
-------------------
|X  B1   |X  A1   | Two RAID 5 partitions (2× 1TB)
-------------------
-------------------
|X  B2   |   A2   | Two RAID 5 partitions (2× 1TB)
-------------------
-------------------
|X  B3   |   A3   | Two RAID 5 partitions (2× 1TB)
-------------------
-------------------
|X  B4   |   A4   | Two RAID 5 partitions (2× 1TB)
-------------------

The X represents where data has to be rewritten and rechecksumed, basically 5/8th of the data.

I can't find specs on the initial Drobo, but the second gen device is specced to have a 500mhz Marvell processor, and that was an upgrade to whatever the first one has. It's going to take that processor a while to recalculate everything.