There is only one preferable engineering solution. Period. A coffer dam around the break on each side and pump. In civil engineering class in my college there there was a saying on the wall: "mud plus mud makes more mud....and you can't do a (censored) thing with mud!". You cannot dump fill in water and create a stable base. Once the area is dry, pilings to rock. Then rip-rap. Then headwalls and one or more large corrugated sectional culverts. Then tamped fill with gravel. I would use earth cement. Portland cement and earth. When close to grade conventional ballast. An alternative would be a pile-driven trestle. A third alternative would be coffer dams, pumps, two 28-day strength concrete wing-wall abutments and a bridge. A steel or pre-cast pre-tensioned concrete span. That should do based on 80,000 pound cars and modern locos. But Canada has an engineering standard. You could go with a truss. Then trackwork. Then remove the coffer dam. Concrete will harden under water but you can't dump it in. There is a caveat. The long approaches may not be stable. So-called Attenberg liquid-limit tests must be performed at frequent intervals in flooded areas. Supersaturated roadbed may survive several passes of a freight, seemingly without harm, and then catastrophically degrade, resulting in a wreck in the middle of nowhere. In any event construction traffic will likely be load-limited. Crews may have to wait till the water subsides and the ground freezes. Or raise and stabilize the approaches. A nightmare on Elm street. In New York this was done by piping in ice-cold saline. Argh!!! More importantly a massive drainage/watershed study is in order. Tree planting, canals and or dams may be in order. A wonderful book on fill is "Moving the Earth" by Nichols - the workbook of excavation. A must read for museum groups with limited budgets and washout issues.
www.garymgreen.com , a modeler, has examples of some actual typical canadian roadbed requirements in appendix six of his site though somewhat dated and blurry.
