I can tell you how we did it:

We performed a huge number of power-up and power-down cycles with a typical AT89S52/MAX1232 application. The microntroller was programmed but not locked, so that we could read the flash code memory after a number of cycles. The result: We never did notice any code corruption with the MAX1232 installed! When the reset chip activates, the micro immediately stops and stays inactived (no ALE pulses) during the whole power-down.

But the situation is entirely different, when disconnecting the MAX1232 and using the simple RC circuit. Many AT89S52 suffer from code changes, but the failure rate isn't constant at all, it enormously differs from micro to micro.

The highest failure rate we observed, when omitting the reset completely.

This made us believe, that not only the power-down but also the power-up is problematic without proper reset.

A sharp power-up and power-down isn't practicable in our applications, because intense supply voltage filtering is used (yes, even for digital Vcc).

Kai