Why does circuits fail when another chip is inserted (different revision - same manufacturer)(diffrent foundry - same manufacturer)(differen derivative)(different manufacturer - same derivative) ?

The obvious: different pinout, different reset polarity, different clock division.

1) FAN OUT. If the data sheet states that a low can be maintained with a maximum current of 4 mA, in most cases the circuit will work with a load of 5mA. To fail you need a combination of an output just making minimum and an input running at maximum. The same goes for the high fan out.

2) TIMING. If a signal is min 40nS, most devices will work even if they require 45nS. To fail you need a combination of an output just making minimum 40 and an input requiring the minimum 45.

3) RESET. The threshold of the reset input is highly variable (see min/max in the data sheet) so a simple RC reset circuit may work in some cases and not in another. I one case of "previously fixed design" we had to specify the capacitor size based on uC manufacturer.

4) DECOUPLING. The noise immunity of uCs i quite variable and insuficient decoupling (no capacitors across uC or the leads to those capacitors being long) can easily make a circuit uC manufacture dependent.

5) more may follow, this is what comes to mind right now.

How to avoid:
1) and 2) : design to the data sheets, not to what "works"
3): use a supervisor chip
4): mount 100nF ceramic and 4.7uF tantalum DIRECTLY across the uC, 100nF ceramic over all other chips (some may need the tantalum as well)

That something works is a matter of design, not test.

Have fun,

Erik