The tRNA matching process is random, and this is exploited by the cell to speed up and slow down polypeptide formation. Most amino acid strings can be derived from any of multiple DNA strings due to synonymous codons. It turns out that the codon usage is not random though, and there is likely evolutionary pressure to use rare codon pairs when you need to produce a protein efficiently: this is called codon pair bias.
One theory is that rare codons will transcribe faster since there tRNAs aren't tied up in other ribosomes producing housekeeping genes, etc.. On the other hand it may be that because it's codon *pairs* that matter, these bits transcribe slower, but give more time for the emerging polypeptide to fold into the "correct" 3D protein confirmation. Either way, there is clearly a significant delay one way or the other.
