Given how well roaming works internationally I think it's safe to say that all technical arguments against national roaming are bunk. The article does highlight the two key points: spectrum without conditions; and decommissioning of sites.
The networks and suppliers are moving towards a classic separation of responsibilities with the suppliers taking over more and more of the business of actually building the physical network – the incentive here is scale and knowhow – which they can then rent to the networks in much the same way that wired telephone exchanges are.
But even with the scale of one physical network for all four operators there are still plenty of sites which are unprofitable. Arguably the consolidation has made the situation even more acute because networks are no longer competing through coverage differentiation. The simple solution is to mandate better geographical coverage to encourage build out and discourage decommissioning. Difficult to do once the spectrum has been awarded but still possible through financial incentives: tax breaks might be best here rather than handouts. The costs associated with the additional cells are pretty easy to calculate so a formula can be derived that works for everyone. This could be augmented with community cells for isolated villages that might want more bandwidth or granular coverage than the networks are willing to provide and you might need to grease BT's palm to lay some fibre here and there for the backhaul.
Long term, however, I suspect that national roaming will come in through the backdoor: 2016 will allow separate contracts for roaming and the rollout of wholesale telephony and data roaming which will make OTT services including VoIP accessible to everyone and will put further downward pressure on margins. But it won't improve coverage unless someone legislates for it.