I'm way late to the party and not a welder but I do know a thing or two about fatigue stress and cracking being a aerospace stress engineer for many years. That said, the first thing I would have done is understand the fatigue stress and how the crack was opening/closing before doing anything other than cleaning up any grease or grim around the crack. Maybe start by jacking up one side or the other would provide a clue if you aren't good at understanding load paths. Not knowing that info... I will say I'm with Mark@Everlast on a gusset style reinforcement plate even one that would only extend inboard some without blocking off access. A larger 'fillet' radius or gusset goes a long way in reducing stress concentrations. I also would not make the repair plate any thicker than the thickness of the adjacent structure.
On fish plates and stress risers, their shape is meant to to gradually change the stress along a structural member. Abrupt changes in cross section means abrupt stress level changes. Another thing that is a good idea to understand... is how you are changing the cross section by adding the repair plate. This is important in bending situations because the added plate can change the neutral axis of the cross section and potentially cause much higher operating stresses on the opposite side of the cross section and causing new problems. This is why it is generally a good idea to make repairs/reinforcements symmetrical if possible. For instance if you were adding a doubler to an 'I' beam you should beef up both top and bottom even if you are only fixing one side. Anyways my :2cents:
