I wish TOCs and government would stop wasting money on wifi; the overwhelming trend is for customers to have their own internet enabled devices. In-train wifi hardware will become obsolete as quickly as the old Virgin headphone sockets for radio channels.
* Wi-Fi doesn't seem like it would be fantastically expensive to install, I strongly suspect it's the ongoing costs that are more significant. So if this hypothetical day comes you can just switch it off and pay no more.
* Phone companies have been dragging their feet on serving roads and railways for years now. There are still major roads and main lines with large patches of missing coverage on major operators. And minor operators --- just forget it.
* Even when there is coverage, a lot of it is GPRS, which today is hopelessly insufficient for modern smartphones.
* Even when there is 3G coverage, modern train windows (especially Voyagers) have an anti-glare coating that attenuates the signal to such an extent that much of the coverage is useless anyway
* At well-served places, the number of people on a train can still cause the cell towers to be swamped. At Waterloo and Clapham Junction when sat on a train I tend to find my phone's internet to be completely unusable.
Many of these problems aren't going to change in the near future, and some of them are nigh-on impossible to solve in the near future. Certainly Wi-Fi will pay for itself in increased productivity even if it's useful for only about five years, and I can guarantee you that it'll be useful for much longer than that.
Now, we could argue that modular solutions are required, where the router, Wi-Fi radio and upstream can all be replaced separately and cheaply using standard protocols, so that when the new mobile broadband standard comes about they can easily be switched to that, and when new 802.11 standards come out they can use that.