As a workaround you could set a downtime for the haproxy backend server checkmk service each time you put it into maint mode.
We stopped monitoring the haproxy backend servers because here they go down and up
due to CI/CD the whole day.
I wrote a special agent and a extra checkmk check that monitors the aggregated backend statistics instead of each backend server on it’s own.
It also does alarming on error ratio ( 5xx vs 2xx http errror codes in percent).