I Do not fully agree. Especially with the example of check_http: The functionality of this check has been extended over the years, resulting in a command line interface that gets harder to map to GUI with each change. Plus there a re bugs like string search: If the string you look for is contained across the border of chunked output, it is not found. There are also various bugs when a proxy should be used.
We could either fork check_http, to like check_http_ng, fixing bugs, cleaning up, breaking compatibility and still have the limitations of an active check.
In case of monitoring devices like web gateways with a full embedded operating system, one might also move to monitoring the firmware version in not so distant future or a REST API might provide better and more performant information than SNMP. In this case an active check is needed anyway. So why not start now?
Going back to the problem in post 11 with the list:
A quick fix might be not calling check_http directly, but creating a wrapper (shell script should be enough) that iterates over the list items, calling check_http for each IP address and then return the output for the IP with the web interface. But I would consider this a hack that is not much quicker to implement than a special agent.