MOTM

> 
> The input bias current into the minus input causes a voltage drop  
> across the input resistors. This would then be amplified and to  
> compensate for this, that resistor is added on plus input to generate  
> an equal voltage on the plus input. Net effect should then be zero at  
> the output.
> 

This is correct but only half the answer. Keep going.............

>> Bonus #2: what "bad effect" results from using it
> 
> It is a carbon resistor, with a 200-250 ppm temp coef. The resistors  
> on the minus input are much better, so you get temperature drift.
> 
>> (but is compensated for later in the design, and what is THAT  
>> compensation?)
> 
> Bit of a guess here: normally you would only need 3300 ppm tempco to  
> compensate the expo converter, but you are using a 3500 ppm tempco  
> resistor. Is that the reason?

No, but a good guess :)

Paul S.