Thanks a lot so far,
now Peter D. threw in some suggestions I´d like to reply to:

> please check:
> 1.
> Was something connected to the SPI ports (P1.5, P1.6, P1.7), which > can generate pulses ?
> Maybe it can help to use pull up resistors.

On P1 I have a couple of keys, connected via 15cm (6 in.) non shielded flat cable, attached ferrite bar, nothing else.

> 2.
> Was the reset circuit right selected for the minimum VCC voltage of > the micro ?

For this application I chose a MAX81 SEURT (3.0V / 200ms) for different reasons.
The app is battery driven and I better don´t use a resetter above
4.1 V. Between that and 3V positive resetters are hard to find

>3.
> Was the crystal working good, e.g. start to oscillate fast enough until > reset was done ?

The crystals in the 5 boards I built in the meantime look allright. The
resetter allows them 200ms to wake up which should be fine.

>4.
>Try to set the lock bits or disable the RCEN fuse to see, if the flash >was erased nevertheless

I did that few days ago, til now no other error has ocurred. But that
somtimes took WEEKS to show up.

>5.
>Was the whole flash erased or only some bytes ?

Only some bytes at different locations.


PS: I just changed the layout a little and built up a new board with no
errors so far. Let´s see. Thanks to Jaceks input I´ll be looking
forward to get the newest revision "T".

For another application I have programmed and used some 1000+
of Atmels 89C2051 (no exclusive reset circuit used).
NOT A SINGLE defective sample or strange behavior in 9 months. Concerning my prototypes I give them a REAL bad time.

Karsten Krause