Use of dye-swaps
Dye-swap replication refers to doing each hybridization twice, the second time with the red-green dye assignment to mRNA samples reversed.
Why? Most cDNA hybridizations exhibit clear dye biases. Normalization attempts to remove these biases. However, these efforts are unlikely to be 100% effective, and so residual biases may accumulate across replicates.
Assigning dyes at random might seem to be an option, but not one which finds much support in the biological community, nor from me. Better to balance, i.e. dye-swap.