Just because 'it worked' with the 89C52 does not necessarily mean your code had no defects. You really need to isolate the problem so you can understand what is happening rather than blame the 89S52. If the problem is consistant, this makes it easier to track down. I would suggest you add diagnostic code is critical places to narrow down the source of the problem. If you think it is code corruption, write code to do a checksum and output the value somehow. Compare this value with the checksum you calculate on the code. Many times I have been struck with similar sounding problems - it is a matter of logically testing and proving things. Eventually you will narrow down to the cause. It might be possible that you have faulty 89S52s, but I would think that a remote possibility if for the most part the code runs.

As for phase control of triacs - start a new thread. There's nothing magic about it though - you apply a current into the gate. To cope with inductive loads, you would normally keep the current applied for the whole duration you want it the device to fire. Depending on the actual triac, you may need a transistor to supply the required gate current.