[GenoMax]

So your paired-end reads are interleaved? Why not use "split -n 2" to divide the original into to two parts. If you have an odd number of fastq records then using an explicit "split -l ((n+1)/2*4)" may be better (n = number of fastq records).