First, a 3 degree curve is nothing compared to the curve in a #9 turnout and even the old liners that had to have the curve manually plotted on graph paper did a good job if the operator was decent. In fact, any section foreman worth his salt can still eyeball line a curve well.
Any of the more modern equipment should give you great track geometry, and unless something developed in the subgrade between the time of the surface and alignment correction and the derailment, I would rule out geometry. You mentioned 90# rail and mentioned a joint in a previous derailment, so I assume it is conventional bolted. If I remember correctly, the tender of the 700 has 2 - 3 axle trucks to distribute the weight of the unit.
A 3 axle truck because of its length steers harder than a 2 axle truck and my guess is those 2 trucks under the tender encountered a joint on the high side with weak ties. They shoved the high side out as they passed over and when the trailing truck on the locomotive approached the joint, the wheels were already crowding the high rail by design, and the low side wheel(s)dropped on the ground.
Hard to tell without being able to look at the site right after the incident, but maybe consider using some oak in the high side joints.
Some time back, one of the managers of my company insisted on using softwood ties in our annual tie program. This included curves of 12 degrees and the following year I was spotting hardwood ties in these curves and replacing every softwood tie in high side joints and shoulders due to the loaded covered hoppers literally ripping the spikes out of the high side joints and causing derailments where the low side wheels would drop on the ground.