RCM3000 hardware problems

Has anyone seen a problem on the RCM3000 core modules where they fail to initialise properly at power-up?
We are seeing it on RCM3110 and RCM3200 at lower temperatures, around 15 to 20 Celsius.
The symptoms are that the timers A and B do not produce the required frequency divisions. Baud rates are wrong.
The symptoms are always cured by taking  /reset_in low manually. That sort of eliminates our code.

Despite the fact that we are buying large quantities of Rabbit modules from Digi they are being particularly un-helpful about this problem. I am afraid that they appear much less helpful than the old Rabbit people.
They have given advice such as "connect an external reset timer to /reset_in "!!. 
After one department suggested that we send a couple of troublesome modules back for them to investigate another department refused to take them because "they are out of warranty".

We have spent many hours testing and trying to eliminate any source of the problem that might be in our application and can find none. We are convinced that it is a design or manufacturing problem in the product.

Due to Digi's hard-line and unwillingness to support their product we have decided to eliminate Rabbit modules or any other Digi product from any short list of components for any new projects.

K.Shail
Sparc Systems Ltd.
UK