Masked ROMs failing, expected lifetime

[Xenon]

Hello all,

In a Tektronix 4051 I have 2 out of 16 ROMs failing (MCM6832). One verifies bad in my programmer, the other verifies good but doesn't function in the 4051.
One dataline of this ROM pulses from 1,5V to 5V. Apparently the programmer says "1,5V is 0" and the 4051 says "1,5V is 1".

Anyway, I have replaced those ROMs with EPROMs and it seems to be functioning again. It crashes from time to time but that can be due to bad connections in my "quick and dirty" ROM adapter.
I have designed an adapter pcb to fit 2732 EPROMs. As soon as I have this fabricated I will know if the crashes remain.

Now I wonder what the expected lifetime of the ROMs is. These are 45 years old.
Here on the forum I read they last about 30 years max. On other websites I read they last as long as any other 74xx logic and EPROMs last only 10 years.
Now what is true?

Will it be wise to replace all ROMS while I'm at it? I have enough used 2732s in stock.
Or is it better to leave as much as possible original (the 4051 is placed in a museum) with the risk of having to open it up again within 6 months or so.

My quick and dirty adapter:

[David Hess]

Operating life for a mask ROM should be essentially indefinite but there have been reliability problems with mask ROMs from companies like MOSTEK.

[floobydust]

What you are seeing, it can be your IC sockets as well. I didn't have great luck with that style of IC socket, fragile and thin metal on the springs and silver black oxides.

[David Hess]

Oh, if those are edge wipe sockets from Texas Instruments, then they are known to be unreliable even on shorter scales of time.

[Xenon]

The high "0" level can also be seen when it is being verified in the programmer. So it is not (only) the socket.

Thanks for confirming these sockets are unreliable. I found it strange to see the contacts on the sharp edges of the ic pin instead of the flat side.
And the contact force is so low the ROM's slide out of the socket from hanging upside-down for 45 years.
This what I saw after opening it:

[David Hess]

The high "0" level can also be seen when it is being verified in the programmer. So it is not (only) the socket.

I do not know how a mask ROM can fail but like I said, some are known to be reliable.  Some fuse based programmable logic suffers from a problem with a similar result where the blown "fuses" regrow.

And the contact force is so low the ROM's slide out of the socket from hanging upside-down for 45 years.

Thermal cycling will do that whether the sockets are upside down or not.  Heat will cause those edge wipe sockets to become loose relatively quickly, and then I always replace them.