I've tried to do nothing else than reseting with Watchdog after my 'boot' seize. It's ok, my application resets but no signal transition on RST pin (pin10 in PLCC) with the scope.
Currently I've only the reset circuit described in the AN461 : a diod, a capacitor, a resistor and a push button.
I suppose the pulse from the micro is clamped due to the capacitor. I'm going to add a resistor as you've mentionned.

Then, when inserting the call to the GoToBoot, I've to manually reset the micro to download the code. So it's half-working but it allows me to recover my bootvector now.

The problem is that I'm crashing too many bootvectors. 3 since 2 hours ! This is really a problem that Philips allowed to put this bootvector in flash. It's not possible to work in these conditions. And It will not allows from myself to use this chip for industrial applications. The reliability isn't acceptable. Bouh, I'm irritated !
Is using the IAP will eliminate these troubles ? I've to evaluate it.

I'm continuing to debug

Thanks
Stephane