How would relabeled chips be able to return the correct chip ID?
Yeah, that would be strange but Chinese do make miracles happen
These chips certainly do not look like more recent revision as on the picture. I see chips from 89 and 92 with apparently different die. Do both of them fail programming the same? If you read it back, does it just program with errors or part of the EPROM remains unprorgammed? Have you tried programming single IC multiple times (set to not do blank check)?
That difference in die size is due to reflections during photography. All the dies look exactly the same. Some parts are shinier than others at different angles. All of them fail like this: 0x200 to 0x27F are corrupted. The first byte is 0x00. Subsequent bytes are 0xFF. Bytes past 0x27F match what I want to put on the chip.
Programming multiple times without an intervening erasure doesn't help.
Programming as a TMS27C256 doesn't help.
Each and every one of these chips fails in the exact same way.
About erasure, yes, I mean with UV. I lit them up for 30 minutes and verified that they were blank prior to programming except for your suggestion that I program multiple times.