Just to finish up this thread...
Ordered a NOS 80C39 made from Phillips dated 93.
Doesn't work.
As another forum remarked in the MicroController section.
And he was correct.
Some versions, such as the OKI Semiconductor MSM80C39, are "improved" with extra features such as additional instructions. Whilst supposedly backward compatible with the original INTEL 80C39 with properly written code, it could behave completely differently if, for example, the code happened to have errors which execute some of the new instructions which were simply ignored by the original version.
National Semiconductors, NEC and AMD, Toshiba (and probably many others) made 80C39s which were functionally identical to the INTEL parts but there is always the possibility of failure due to small differences in power consumption, I/O drive capabilities, input thresholds, reset level threshold etc. Also crystal oscillator variances may mean that not all crystals are guaranteed to work with all the different manufacturer's parts. If the circuit was properly designed then it will likely work fine with any of them but there is no guarantee. We fell foul of this with the bus port initializing into the input state after reset - but not before initially powering up in an output state on one manufacturers version but not on another, causing some to self destruct after enough power cycles.
If it is for a critical application or you only need a few I would try to get another ZyMOS part with roughly the same date code (47th week 1987) - later parts may have been manufactured with a mask shrink which would change the parts characteristics slightly - or even fixed some errata (bugs) which the code happens to, inadvertently, rely on. That will probably be difficult to source though. You might get lucky if someone here happens to know of any equipment which used those ZyMOS parts which might be easier to find than the bare chip.