Farnell have 35 different makes of 93C46 in stock!
I can vouch for strange things when they are failing, for example corruption of certain parameters but not others. I have to change them regularly in a particular type of lab instrument, which might accept a full calibration only to find the next day it's changed a parameter that should be, say '100' to something like '--Dh' or other nonsense.
These chips have a limited number of write cycles before they go intermittent.
+1 on the plcc socket too. I eject the chip and lightly sand the chip pins with fine wet & dry before replacing. It's cured a few weird faults.
http://uk.farnell.com/search?st=*93c46*