Inconsistent species results

[mcleanlab]

Hello,
We have been using Midas for some time, specifically a docker version
quay.io/fhcrc-microbiome/midas
In setting up a new machine I installed Docker and did a test run of Midas using some previously analyzed data. The new results did not match the old results. The pan-genomes were the same, but the number of hits was different. At first I assumed that this was because of a database update. However, I found that I can run Midas species on the same dataset, with the same database, from the same docker image just 20 minutes apart or so and I get different results.
I have not previously noticed this behavior, but haven't looked for it. I had assumed that the matching function did not involve randomness and that the results would be repeatable. Is Midas working as intended or has something gone wrong with my implementation.
Thank you for your time
Mcleanlab