The RSEM utility rsem-calculate-expression has a setting --sort-bam-by-read-name. The documentation states that doing so will result in deterministic maximum likelihood estimates, at the cost of longer run-times and larger memory requirements. By default, this setting is disabled

Given the same set of input data files (reads and RSEM Reference) does RSEM calculate the same Estimated Counts / TPM / FPKM over multiple repeated runs without this setting enabled?

I am currently running a repeat on my data set (using default RSEM Settings) to compare A vs. B; I was hoping someone with RSEM experience could comment on how this setting impacts RSEM results.