Something to consider when testing, or using, "hidden" sites and pages.
I'm genuinely surprised that it isn't causing at least minor problems with some sites for more technical users. I manage a virus-scanning web proxy for our company and this is enough to throw up regular little glitches with all sorts of sites which "just work" when going direct.
What are the glitches you're seeing?
You're probably right that this system isn't proxying everything, so its impact would be a lot lower. It all depends on what layer it tries to operate, which we don't really know. I'd love to see an investigation from someone on TalkTalk.
Of course it hasn't; It's only been 2 years. You won't notice unless you're trying to build or deploy something innovative that breaks some of the assumptions that the designers of the filters never anticipated.
Layered protocols exist to allow them to evolve over time with a minimum of complexity (and cost). When you break the layering, you make that evolution more expensive. Look at how expensive it is to write a basic app that has peer-to-peer conectivity, thanks to NAT.
How are we supposed to deploy the next WebSockets when there are transparent HTTP proxies everywhere that aren't aware of it?
Economically speaking, widespread deployment of these sorts of filtering technologies are exactly like pollution: they look fine at first, but eventually they end up being very expensive for everyone, even the people who never used them.
Whether you agree with it, like it, or want it, is an entirely different question (I don't want it), but it hasn't caused significant issues yet.