As far as I know, the preset system still operates essentially the way you have described it: If you put the name of a Preset in the the DSP field for a file, MC will load that preset and apply it. MC will keep that preset applied until you manually make changes, or until it sees another file with a preset in the DSP field.
So, in practice, if you want to use per file DSP, you have to have a DSP value for all files. The normal advice is to develop an every day DSP preset as a Base. This would have all of your normal Output Format settings and anything else that you always want to use. Then, when you are developing your per file DSP settings, you use the Base as the starting point. Then do the parametric (or graphic) EQ how you want it, and save this as a per song preset.
Then, you apply the Base preset to ALL of your files, by setting the [DSP] field to Base for all of your files. As you said, this means that you have to tag ALL of your files. Next, you tag any of your files with individual DSP presets so that they sound different than the rest (or are corrected in a different way).
This is how it works right now. I find it to be too cumbersome, so I don't use it. Plus, per file DSP is kind of a rabbit hole in my opinion. I don't want to be endlessly tweaking the EQ settings for various songs. It's pretty cool, but I think I personally might waste a bunch of time messing with it. (thus the rabbit hole comment).
It's a usable system if you follow the process above; its just not for me.
Brian.
