Track Design and Notes
Although derailments on model trains aren't as serious as those on real trains they can still be a major hassle and even cause injury. Some people fit skid bars to locomotives to reduce the chance of damage. Locomotive derailments might be due to track faults or obstructions (sticks or stones) while carriage derailments might be due to track faults or passengers moving around and upsetting the carriage balance.
Most rails are 25 by 10 or 20 by 10 rectangular steel bar. Wise folk always position the bar so the top inner arris is rounded. These arrises should have a radius of about 0.75mm. This radius must be less than the flange root radius on wheels to reduce drag when the flange is guiding the wheel. Most track has the rails welded to steel sleepers (30 by 5) and this is great because it maintains the gauge and lasts for ages. But if anything hits the sleepers it will probably bend or break something.
5" gauge track in Australia is complicated by the different wheelset profiles in use here and sometimes by the need to combine with 7.25" gauge track either by crossings or sharing a common rail.
The pages here are PDF copies to avoid the overhead of maintaining both the original files and HTML copies.
I haven't found any specifications limiting the track twist rate (change in superevelation) or vertical curvature (rate of gradient change). Both these rates affect the weight distribution and can result in wheels lifting or riding over the rail and derailing. Both are a vertical deviation from a straight line and consist of a smooth change and short abrupt changes. A gradient change is when both rails deviate equally. A superelevation or cant change is when the rails deviate unequally.
A gradient change will be either a crest or a valley. Crests cause problems for locomotives particularly those with leading bogies. Crests on curves are even worse.
The weight distribution imbalance from a superelevation change will depend on the rate of change and the length of the vehicle. Unloaded passenger carriages will probably be the limit here because of their long wheelbase and firm springing required for their use when loaded. The rate of change here must be less than a gradient change because it is much easier for a wheel to lift due to a superevelation change than both wheels on an axle to lift due to a crest.
Considering the above limits the worst case will be a valley in the outer rail of a curve when there is no similar valley in the inner rail.