Hallo Karthik,

there are many possible reasons for your failures. Have you
checked your stack performance? Is there an overflow at certain
conditions?
Another point could be the loss of Faraday cage: If your applica-
tion does not sit inside a closed metallic enclosure static dis-
charge can flow through some wiring of your application, e.g.
across your PCB. ESD spike overlayed to normal signals can do
both:
Let your micro hang-up and erase your flash memory.
Also 'normal' noise on power supply or any other signal line can
lead to your failure. Please keep in mind that circuitry of micro-
controller is very very complex and even a very short spike at the
wrong place and the wrong time can destroy highly dynamical perfor-
mance of microcontroller.
By the way, do you have a good reset controller on your board?
So, seems to become a long night for you, hunting for spikes, over-
voltages, surges and dips with the help of a good digital oscillos-
cope.

Sometimes wiring managment is cause of strange failures: Many people
avoid making local earth connections in the hope that floating
grounds will help them preventing problems, but if there are esd
events, bursts or surges they see no direct way to earth but travel
along whole wiring and make trouble far away from the point where
they were 'borne'. To avoid this whole application should be sur-
rounded by a complete Faraday cage. At any module there should only
be one point where cables come in and go out. Never allow that such
currents flow accross PCB. All cable screens must be connected at
a high frequency backplane at the point of cable entry.
If screening of cables or Faraday cage cannot be used heavy filte-
ring is a must. But even with this measure sporadical eventing
failures are difficult to prevent. So, do always use screened cables
and Faraday cage, even when it seems to look uncool or people are
'laughing' about you.

Good luck,
Kai