Now that we all know what Erik does NOT make (guys, he definetly does not make toys!!!), I guess what does he make...
;-)
we make a sign and announciator system for municipal buses that based on GPS announce next stop, change the signs, let the supervisor in his office see where the buses are, estimate time of arrival and display that at the shelters ............. In other words, leave the driver to the driving.

I agree with you, Erik, anyway: this tecnique of using a single clock source for two micros is not good.
I disagree, the ONLY way to run multiple micros is with ONE clock source. If "this technique" refer to using XTAL2 you are right.

Maybe this technique is not the best of the world, but it worked with the AT89S53, AT89S8252, and maybe it works with any other 8051, but it does not work with AT89S8253.
This is often seen when one micro is used outside the guaranteed parametres. If upgraded to a newer chip or even the manufacturer change foundry everything fall apart. In your case I guess that since most often newer chips are faster (in risetime) that is what got you. Asd Kai stated it is next to impossible to measure oscillator output.

BTW, also using TF2 to generate a software interrupt is not exactly the right way to use it... I know.
why not, I have done it many times.

Erik