I'm using version 4.7. (Not sure about the category I put this bug in.)
As Michael pointed out on the mailinglist (subject "Dropbear commands not executed on resource"), commands on a resource are not executed properly, when the OpenQRM server was installed with a different prefix /usr/local/openqrm for example.
This happens because the command submitted via dropbear uses the $OPENQRM_SERVER_BASE_DIR to also build the command line on the resource
/usr/local/openqrm/bin/dbclient -K 10 -y -i /usr/local/openqrm/etc/dropbear/dropbear_rsa_host_key -p
1667 email@example.com "/usr/local/openqrm/bin/openqrm-cmd reboot"
If the server image of the resource is (OpenQRM-wise) untouched, the OpenQRM binaries will be located in /usr/share/openqrm.
Michael put a patch in his posting which fixes this behaviour in sbin/openqrm-exec (I attached Michaels patch). Yesterday I stumbled across another problem when using KVM. But I don't exactly remember where it happend. My solution to this problem was to create a link inside the server image from /usr/local/openqrm to /usr/share/openqrm.
So a global solution could be to create a link inside the image, whenever $OPENQRM_SERVER_BASE_DIR is not /usr/share/openqrm.