It means whatever you'd like it to. That is, you decide when a given model is "good enough" that you're happy with it.
We all have to decide how much time, effort, and money to put into whatever we're modeling. In addition, our own skill level may be a limitation, although we always have the option of paying somebody else to do what we feel is beyond our skills. For example, hiring Pierre Oliver to build your resin kits. Many people pay to have decoders installed for them. I know a lot of folks that have paid others to finish their brass steam too.
Many modelers have different bars for different things. I've seen layouts with exquisitely detailed scenery and/or models, with sloppily applied ballast. It's not uncommon for modelers to put more effort into locomotives, cabooses, and home road cars than those or foreign roads. Although I think this often has to do with the information readily available, although the internet has changed that considerably.
For me, "good enough" is a moving target. Many of the compromises I choose today are "temporary" and I intend to address them in the future. For example, for most RTR cars I just put them on the layout as is. And for plastic kits, I often do the same with the idea that if I had purchased them RTR, they would have just gone on the layout. That doesn't mean I won't go back and make modifications, add missing parts, etc. Just that I'm not doing so now.
This also applies to things that are beyond my skill set right now. As my skills improve, I'll go back and upgrade things at a later date.
Other times, projects go on hold while I accumulate the information or parts I need. This happens because I often have more opportunities for research than hands-on modeling. Chris and I joke that the curse of prototype modeling is that once you know, you know. I'd rather not know that K-1-d #479 has Southern valve gear, and that it was the locomotive assigned to the New Hartford local. But I know. And I care...for now.
When I get to the point that I'm getting my steam ready for operation, the model very well may retain the Baker valve gear that's already on it. At least until I, or somebody I can pay, can replace it. It might never be altered.
Because the reality is, I'd rather know and choose not to make the modification, than not know.
My goal, though, is for everything to be modeled as prototypically as possible. The trains, the track, the scenery, and also things like the movement of trains, operations, and the paperwork too. I've been going over some of the operations and paperwork recently, but what is "good enough" for my physical models?