Different final reads after pre-processing with same code running twice

Hi, I run the pre-processing procedures twice using the same code for the same data (I deleted processed fastq data accidentally and need to re-run everything). I found that the final reads are different from the two different runs. Is there any parameter that is generated randomly in the kneaddata? I used “–threads 8 --local-jobs 8” in each run. Could the different reads be caused by parallel setting?