Imagine if, in 2002, the cable companies had come offered a little, Roku-esque box which could hold two almost DVD-quality movies. A consumer would have attached a Y-splitter in front of his existing cable box and then added this new box in parallel. The box then would have connected to the TV's analog inputs. One would have loaded up his box by going to the cable company's website and picking movies from a large catalog -- one movie for slot "A" and another for slot "B". Each slot would have taken 8 hours to download over the cable line and then could have been watched using a minimalist remote to pause, rewind, etc. -- just like a VCR. Consumers could have gone to their local cable company office and picked up this box along with unlimited movie service for, say, $20/month.
I've purposefully suggested the crappiest, most minimalist implementation I could imagine as a thought experiment. Would this have been sufficient to combat Netflix? Presuming identical catalogs, why would I fiddle with snail mail, scratched-up DVDs, etc.? Such an offering certainly would have created an uphill battle for Netflix.
It is not as if video-on-demand services had not been discussed since at least the early 90s. What I have proposed is worse than what was suggested even back then, right? So, why didn't the cable operators do anything? I'm bet it had to do with fear of cannibalizing existing revenue streams. Why pay for HBO (sans original content of course) if you could pick ANY movie? How many industries have hurt themselves by not cannibalizing their existing revenue streams with sufficient aggressiveness?