On the railways in general there is a strong inclination towards keeping what works as the potential downsides of getting it wrong can be literally catastrophic

On the railways in general there is a strong inclination towards keeping what works as the potential downsides of getting it wrong can be literally catastrophic. Things like solid-state interlocks, are reliable and trusted; it's taken a lot of work to develop the same kind of trust in software based systems. I worked in railway safety back in the early 2000s and moving block signalling was like nuclear fusion, the great hope for increasing capacity and always just around the corner. While it's in use on some underground lines, it still isn't in widespread use on surface lines, it's just proved really hard to get right. We are getting there, it will form part of future systems in Europe, but it's taken decades to get to the level of maturity where people will trust it to carry thousands of people at intercity speeds through complex rail networks.