I didn’t know about tunnel diode before this October in supporting this post (How to generate piece-wise linear multi-peak resonant tunneling diode model - QSPICE - Qorvo Tech Forum)
Well… I can understand why this I-V relationship can pose a challenge, as a single current value can result in multiple voltage solutions. If, in any case, the simulation attempts to determine the node voltage from the current value, the solving process may jump between different solutions and be unable to converge on a solution, resulting in a timestep that is too small.
Anyway, I believe it is very challenging to integrate this I-V relationship into a complex circuit. Having a single V-source to obtain its I-V is simple. However, anything more complex may become entangled in its characteristics. Just my thoughts on this topic.