I have been following this thread and finally decided to stick my nose in. I agree with Russell Bull that EMI related issues can come from any direction and add up.
I use an AT89C51IC2 in my design and the WD functionality works just fine for most things. I work in an EMI noisy environment which include DC Motors and their related relay switches and 100W Class D amp in close proximity. Now and then the micro would either latch-up or do something screwy when I kicked on the motors to raise the platform. As it turned out the SPI and internal UART ports were picking up a huge amount of radiated energy. My bad, I forgot to use transient suppression diodes on the transmit and receive lines. But the point to all this is I use a 2 layer board, which I mill in the shop, with no ill effects and I use every available comm port on the device (RS232, SPI, and I2C) along with (2)8255 I/O expanders, a MIC preamp and 2 digital pots for volume control on this 4x6" PCB. Techniques I learned from a book called "Grounding and Shielding: Circuits and Interference" by Ralph Morrison opened my eyes to a lot of things I was doing wrong. It also helped me improve the EMI performance in the design of the Class D Amp sitting about 8 inches from my micro board. This system has been running successfully 24/7 for 3+ years.