And amusingly, the bricking worked because again the clones worked better. The clones had implemented FTDI's EEPROM interface as designed, 16 bits at a time, while FTDI apparently had used a 32bit array for the internal FT232 EEPROM, and so would buffer 16 bits and only write all 32 bits when you issue writes to odd addresses. FTDI's bricking code uses writes to even addresses only (including computing a preimage attack on their checksum algorithm, as the checksum is at an odd address so they can't change it), which is why it doesn't affect the real ones.