Certainly all of this would be completely impractical if it required a "parallel thread of consciousness" (where did you get that phrase). Thankfully recognizing situations isn't something that has to take conscious attention. If I'm walking down the street and someone I pass is wearing a gorilla costume that fact will be promoted to my conscious attention by other parts of my brain without me having to consciously think to myself "is this a gorilla" whenever I see someone. The idea here is to train yourself to notice when, say, you're confused the same way you'd notice a gorilla costume and only then deal with the issue consciously.

Of course automatic recognition isn't perfect as the video you're referring to shows. But recognizing something most of the time is far superior to never recognizing it.

And I think a better model for the human brain is a hard core with a single thread embedded in a much larger FPGA fabric that can issue it interrupts. Only a tiny fraction of the sense data that enters our brain actually impinged on our conscious awareness and the directives that our consciousness issues such as "pick up that cup" are multiplied hugely in complexity as they are translated into precisely calibrated exertion rates for dozens of separate muscles over time.