If I’m reading marcan_42’s comment correctly, the FTDI chip defers any EEPROM write request to an even address until it receives the next write request to the subsequent, odd, location. If the next write is somewhere else, it discards the original data.

The 3rd-party chips, on the other hand, perform each write immedately per the official documentation. The malicious driver uses this difference to overwrite the firmware of the 3rd-party chips with carefully-crafted nonsense that has the same checksum as the legitimate firmware.