Well the signalling system does need on overhaul. There's no question that current systems are not only expensive but out of date - the methods going back to the 1920s/30s.
The power boxes of the 60s were like an early invention of the internet, with electro-mechanical computers (aka relay rooms) communicating with the main interlock via a telephone cable.
Next came solid state interlocking in the 1980s, computers doing the logic, which by today's standards is comparatively simple. These computers ran (and still run) at 1 mhz, similar to the home computers of the 80s. The latest ones run at 2 mhz, such as Manchester South. Most of the reason is to ensure safety as only a few microprocessors can be assured to deliver constant results. Empty SSI interlocking cubicles cost £11,000, with prices of modules a few hundred pounds, I guess it all adds up. A price list can be found below.
http://www.signallingsolutions.info/pdfs/brochure.pdf
Today we have wifi, high speed high broadband phones, GPS - and that's just in a mobile phone which costs £150. I'm not an engineer, so I don't know why current railway signalling systems need lights on poles, cables to every cubicle and so on. But it strikes me this is an area where innovation is much needed. Ideally you would have a common system which is effectively portable and modular, the trains could either run in automatic or be driven on the main lines and be cheap enough to fit to all trains (including steam and enthusiast specials) without issue. The whole world should also be adopting the framework for it, so trains can run anywhere.
As for £35 quid per train Death, I think the latest ECTS fittings per train are around £400,000 each. (I have no idea why, I'm not an engineer). Line speed is also limited by track curvature, you would only get 20% extra with tilt.