I think the discrepancy is due to it still running within a container, this time from within the homeassistant container and not the SSH addon.
I’ve did some further testing and able to now pull stats from HAOS itself allowing more accurate resource monitoring.
I copied the check_mk_agent to /mnt/data/ within HAOS however when running, it produced errors relating to illegal ‘-f’ option for some export commands. I removed the -f flag from the following lines within the check_mk_agent (lines 381-383) which allows it to run successfully:
export -f run_mrpe
export -f waitmax
export -f run_cached
After updating the Individual program call instead of agent access command line as below, I was able to get more accurate stats:
ssh -p 22222 root@$HOSTADDRESS$ /mnt/data/check_mk_agent
This also updated the hostlabels within Check_MK to vm from containerized
It could be that once you do this, the two Memory values will match.