BTW has ATSC 3 learned anything from the idiocy of the current scheme? Do we get a system that's actually designed by people who know what they are doing?
eg
That looks like about the best one could hope for. Unicode support is there. A separate stream is there. Like everything WWW based, there's massive text overhead -- but that's surely swamped by the AV data so we can live with it.
The one thing that isn't clarified is whether the stream can be encoded with richer FEC. (In a perfect world each of text, audio, video vectors, and video texture would be encoded in a separate stream, each with successively weaker FEC. In practice there appear to be so few people familiar with all the relevant layers of the stack that this always slips through the cracks 
)
Captions on internet streaming content mostly work pretty well now (as opposed to, say, five years ago) so hopefully we'll get that level of quality. The one problem I still see too often with internet content is terrible sync between the captions and the AV content (not just a few frames, which is dumb but can be lived with) but off by 30 seconds to many minutes. I guess you can't completely hide from incompetence, via legislation or via technology.