Lots of reasons:
- Better value
Only resistors of a particular size and rating could be found, but not in the required value. So they were connected in parallel to get the correct value and rating.
- Lower inductance
Wider trace width reduces inductance ~proportionally. Less overshoot on the current sense signal.
- Greater power dissipation
Again, resistors of a particular size could be found, but not in the required power ratings. Power resistors are also physically larger, so exhibit more inductance.
- Better tolerance
If all the resistors are independent random variables within their tolerance range (say 1%), then N resistors in parallel exhibit sqrt(N) times tighter tolerance. Doesn't matter much for N=2, but up around 5-20, it's not too bad (brings it down to ~0.25%, say).
- More flexibility
Probably a big driver on a dev board: if you need some oddball value for your test, you can assemble it from two (maybe more) resistors in combination. The other reasons might not matter at all, it's just convenient.
A lot of commercial equipment will do it for the first two reasons, because through hole resistors have higher power ratings and lower cost (or did have lower cost; that's changing more and more these days), but higher inductance. Availability of very small values may also be limited, although if we're talking production, I'd be surprised if one couldn't convince any resistor manufacturer to spin a million of some oddball custom value for peanuts.
Tim