The required minimum runway width obviously varies depending on how close to the centreline the aircraft is when the engine fails. Assuming that aircraft wander patterns are normally distributed, based on field measurements (Brown, D.N. and Thompson, O.O. (1973). "Lateral Distribution of Aircraft Traffic,"Miscellaneous Paper S-73-56, U.S. Army Corps of Engineers, Waterways Experiment Station, Vicksburg, MS.), the distribution of aircraft on runways is assumed to have a standard deviation of 1.55 m. So at the 95% confidence level, this is mean + 2 s.d = 3.1m. Aeroengineer1 had used 2m (which is the 80% confidence level - also a reasonable assumption). Either 2 or 3m could be used, and I now think that the 5m value used in my earlier post was too conservative.
Aeroengineer1 calculated the Boeing 737-800 outer wheels to be 0.36m inside the runway for the 30m wide runway and a 9m lateral deviation. That is at the 80% confidence level. At the 95% confidence level, the outer tyre would be 0.74 m outside the runway (and onto the runway shoulder).
The Boeing way of handling this seems to be to reduce the takeoff thrust, and load to more forward CG to improve directional control. This would reduce the lateral deviation to less than 9m, and so even at the 95% confidence level, should keep the aircraft on the 30m wide runway, albeit very close to the edge.
When the wheel is on the runway edge, the load and load spreading will be taken partly by the runway and partly by the material/subgrade in the shoulders. Since the number of aircraft ever doing this is hopefully very few, we need only design for a few passes. It should be OK, provided the shoulder pavement material is kept dry. From the highway concepts of edge wetting and shoulders, we know that a sealed shoulder of 1,5m protects the edge of the carriageway from the edge wetting effects of rainfall. So to be sure about 737 operations on a 30m wide runway, we might look at something like a 1.5m sealed shoulder with a bearing strength suited to a couple of passes of the aircraft.
The engine is hanging out a bit further than the wheels, and is well out over the shoulder. But then it is probably the engine that failed, causing the yaw in the first place. So there may not be too much risk of FOD.
This accords with slide 18 of Root's Boeing presentation:
No regulatory link between VMCG definition and actual runway width, so the maximum 30ft deviation could result in reduced (or nonexistent) clearance between outboard main landing gear tire[s] and runway edge…
and his slide 30:
Increased risk of Foreign Object Damage (FOD) to wing-mounted engines
The real trick now is get the 747 onto the 30m runway -
here's how it is done