I've been working on LDraw programs (with pauses) for over 10 years now, and in my experience it's best to hope for the best but prepare for the worst 
Also if more programs apply smoothing some of the parts will eventually 'automatically' get fixed by the part enthusiasts to make them look better in their favorite tool. So over compensating might actually be a bad thing.
Currently I tend to do corrections as far 'normal' performance allows it, In my older renderer i used to disk cache the unique point preparations, while LDCad does everything real-time using multiple cores now. The is the direct result of PC speed progress, fact it can be done fast enough in real-time, made me want to add smoothing in the first place.
So in short:
#1: It's largely up to the community I think, in the meantime tools need to compensate to the amount authors find doable/necessary.
#2: I'm sure any well tested mass correction / authoring tools would be welcomed by the library admins / part authors.
#3: I do think a smoothing (bfc like) direction meta could make useful addition to the spec, but it needs to be thought out deeply in order to decide if it's worth including. So it could take some time until it's available to the authors of both parts and tools. On the other hand the problem areas seem to be mostly detailed curved areas, something that can also be addressed using the new texture extension. So maybe it isn't even worth to fix the 'past' I haven't tried including the hard edges while smoothing yet though, so there might be some grounds for useful meta direction with those.
Also if more programs apply smoothing some of the parts will eventually 'automatically' get fixed by the part enthusiasts to make them look better in their favorite tool. So over compensating might actually be a bad thing.
Currently I tend to do corrections as far 'normal' performance allows it, In my older renderer i used to disk cache the unique point preparations, while LDCad does everything real-time using multiple cores now. The is the direct result of PC speed progress, fact it can be done fast enough in real-time, made me want to add smoothing in the first place.
So in short:
#1: It's largely up to the community I think, in the meantime tools need to compensate to the amount authors find doable/necessary.
#2: I'm sure any well tested mass correction / authoring tools would be welcomed by the library admins / part authors.
#3: I do think a smoothing (bfc like) direction meta could make useful addition to the spec, but it needs to be thought out deeply in order to decide if it's worth including. So it could take some time until it's available to the authors of both parts and tools. On the other hand the problem areas seem to be mostly detailed curved areas, something that can also be addressed using the new texture extension. So maybe it isn't even worth to fix the 'past'
I haven't tried including the hard edges while smoothing yet though, so there might be some grounds for useful meta direction with those.