Christopher Rackauckas

That sounds about right. Odd it takes so much though: MTK only needs like one or two tie-in points, and so does LabelledArrays

I think that's a good idea for the interface. We'll need to specifically setup methods for it, and I think it only gets rid of kernels in Shu-Osher form, so...

Yeah this is an odd issue. The issue is that the default `dt` happens to be `1e-20` away from the `tstop` that is generated for the callback, so it steps...

It's another instance of https://github.com/SciML/DifferentialEquations.jl/issues/570

>but at 1e-16 it reappears. `1e-16 < eps(Float64)`, so it's below dtmin.

It looks like you're not using an ODEProblem but a DynamicalODEProblem of sorts? Like a SecondOrderODEProblem?

It looks like this is a bug in the stiff solver handling of ArrayPartitions. If you code the 1st order system yourself it should work, but the automated handling likely...

Fixed by https://github.com/SciML/DiffEqBase.jl/pull/536

>I have also tried using the PositiveDomain callback and isoutofdomain functions but both are not able to strictly enforces a positive solution for u2 Is it an instability or does...

Summing up from the Slack chat, it seems like `AutoTsit5(Rosenbrock23(chunk_size=8))` and `AutoTsit5(Rosenbrock23(chunk_size=4))` both reduce the compile time by half, and going to `AutoTsit5(Rosenbrock23(chunk_size=1))` reduces the compile time down to the...