Kei,
I am a newcomer here, but in my loong life I have been working with 8051 (with external EPROM and RAM), 87C52, Phytec modules, DS89C420 (wonderful chip), AT89C52/S52 and S8252 (among others), in small and medium size projects, using as reset RC, MAX1232, DS1233, pushbuttons, pedals, etc ;). Also in very noisy environments as PWM motor control, for example. I have never seen such a weird behavior as in AT89S8253.
Kei said:

If you check carefully
Atmel AT89S8253 manual
page 11, you will see Internal RESET signal (fig. 7-1, bottom), that is not not related with RST pin 9.
This signal is the REAL reset , that resets ALL the internal things. Pin 9 RST only resets PC, ports and SFRs. So, no matter what do you do to pin 9, if Vcc rise time is not fast enough.
I think that RST works totally diferent than AT89S52 (your example)

So, my initial comment, for people like me, who are stuck for ANY reason to AT89S8253, was check these non documented issues :
- Use a clean Vcc. This chip is more delicate than 8252
- Check that Vcc rise time is fast (<1ms). Otherwise, you get mad.
- If you access internal EEPROM, do not use ALE as clock for any other device, in spite of you previous experience.
On the other hand
- It is a non-expensive chip.
- Easy and fast to program (I modified my parallel programmer quite fast)
- Has an x2 mode (that I did not fully test) that means 6 oscillator periods per machine cycle.
- Fast EEPROM page writing (32 bytes in 4ms).
- Good internal reset (if Vcc rise time is fast enough…)

That’s all.
Daniel (56)