What you recommend to monitor very old systems like Debian 5 or Ubuntu 6.06 up to 14.04?
In addition to checks such as http, ping and ssh, local tests like apt, disk, procs, load, mailq are supposed to run on every host.
Newer systems such as Ubuntu 18.04 and Windows Server 2016 are also in use and are currently being monitored through icinga2 agents and config sync. The configuration on the master is done via the Director module. About 500 hosts should be monitored.
You could use snmp or the (evil) nrpe
“I’m a bad boy!”
What about by_ssh? Are there any restrictions compared to the icinga2 agent and config sync?
I can only recommend SSH being a secure alternative. NRPE is strongly discouraged.
How to set “vars.by_ssh_command” for an service template with Director?
Why it is not available via External Command > by_ssh > Fields?
…Or is there a fundamentally different approach?
Whenever fields are not available via the external commands import, you can define them as data fields in the infrastructure section. See here for an example: Bit of a predicament with commands on Linux clients
Debian 5 , ubuntu 6.x and security 
Even if they are really old, if they are run in isolated environments, I wouldn’t bother with exposing too much into the monitoring. SSH is still the safest bridge in this regard in contrast to NRPE.