[cctalk] VT100: Failing 2114 Chip Replaced With One With The Same Fault

I posted recently that I had identified a faulty RAM chip in my VT100 and having replaced it the terminal seemed to get further into the self test. After further analysis with the logic analyser I have realised that it is still failing the RAM test, despite an apparent change in behaviour. I think the change in behaviour could be simply due to the slightly suspect keyboard cable. But here is the puzzle. When I first identified the faulty RAM chip if found that address 0x2408 would read back as 0x0A instead of 0xAA. I reckon this equates to E50 in the schematic, as it is the upper nibble of the second bank of RAM. I replaced the chip with one I bought recently. It turns out the self-test is still failing at the SAME address. I find this really hard to explain. It can't be the chip selection logic because then the addresses 0x2400-0x2407 would also fail and I checked the CS signal with the logic analyser just to be sure. I also checked the address lines directly on the RAM chip for any stuck bits and they seemed fine too. What are the chances of two 2114 chips failing at exactly the same address? Is there some failure mode I might not be considering? Thanks Rob