I don't know and I do not use ODE but its sounds like a major problem. Presumably it would need to determine how coplanar the other triangle is that shares this edge at point of contact and adjust the response accordingly. Sounds easy but it probably isn't, not if you want to do it quickly anyway.
I am suprised ODE has this problem though, after all Blitz seems to manage its sphere to mesh collisions without any issues of this sort, that I have noticed anyway.
Maybe this quote from the ODE docs suggests some awareness of this limitation
In real life, contact between two bodies is a complex thing. Representing contacts by contact points is only an approximation. Contact "patches" or "surfaces" might be more physically accurate, but representing these things in high speed simulation software is a challenge.
Probably also worth looking at 10.7.6. in the ODE docs which makes further suggestions for improving trimesh collisions, such as reducing the stepsize. Not sure if that would be applicable to your sphere/cylinder case or not.
I am suprised ODE has this problem though, after all Blitz seems to manage its sphere to mesh collisions without any issues of this sort, that I have noticed anyway.
Maybe this quote from the ODE docs suggests some awareness of this limitation
In real life, contact between two bodies is a complex thing. Representing contacts by contact points is only an approximation. Contact "patches" or "surfaces" might be more physically accurate, but representing these things in high speed simulation software is a challenge.
Probably also worth looking at 10.7.6. in the ODE docs which makes further suggestions for improving trimesh collisions, such as reducing the stepsize. Not sure if that would be applicable to your sphere/cylinder case or not.