Testing is a huge part. Learning how to test effectively, how to spot missing or wrong data, having multiple machines to test on, all helpful.
The bug in 3.3.17 for example happened because all my local systems, and all my remote systems, did not trigger that bug. But it did trigger for other people, including the packager, who tested it, luckily.
The most helpful long term is a somewhat committed level, where it's ongoing and I don't have to do anything. Big public tip of the hat to mr mazda for that type of long term behind the scenes interest and suggestions and ideas. When you look at -G today, you are looking in many ways at his ideas and suggestions being brought to life.
Extremely helpful things are generally software related, like a new wm, new desktop, how to get data off of it, version info, etc, compositors in wayland, new display managers, all that stuff. The sofftware stuff is the hardest because I can't ever guess at how version data works, or how to detect a compositor or window manager or desktop, it is always by a series of tests, and I have to know which test detects it.
Wayland compositors are a big weak spot, inxi can't get data from gnome or kde wayland at all yet, for example, no documentation, no data, nothing. new wayland compositors pop up constantly, I generally only learn about them at random, or by actually checking a 'best 40 compositors in linux for 2022' online list or something like that.
It basically comes down to stuff you care about, and find interesting, that's the stuff most likely to hold long term attention. For me, I have to force myself to care about whatever I'm working on this moment, and once it's done and stable, I forget all about it more or less, so that care is only short term.