node_systemd_unit_state metrics silently ignores units that systemd failed to load

Systemd units have two status fields, "load state" and "active state". Node Exporter only reports on active state. If a unit has any load state other than "loaded" it is silently ignored (unless debug logging is enabled, but it still does not appear in metrics).

Practical example: if a .timer unit has an invalid OnCalendar value (if like me you mistakenly assumed it would support typical crontab syntax), then the timer will fail to load and never be reported in any metrics by Node Exporter.

This filtering is done here: https://github.com/prometheus/node_exporter/blob/2179f0a34d2d7b6212f3a1c647d5aca44ffa33e5/collector/systemd_linux.go#L494-L499

It was added in this commit: https://github.com/prometheus/node_exporter/commit/0fdc089187c3d5d9fe62e528653986ee6e71325b

I am not sure why this filter was added, but i feel that being able to alert on units that failed to load would be a useful safety feature. Could this filter either be removed or controlled by a CLI flag?
Thanks!

Start in collector/systemd_linux.go at the filtering logic around lines 494-499, then read the linked commit that introduced it. Determine whether failed-to-load units should be reported directly or controlled by a CLI flag, and verify that the resulting metrics include units with non-loaded load states.