Sam Isaacson

If it helps in debugging this, the problem seems to go away if one doesn't use `prob_func`.

@TorkelE this seems to fix it for me, does it work for you? ```julia ejprob = EnsembleProblem(jprob,prob_func=(p,i,r)->remake(p), safetycopy=false) ``` Looks like the problem is `JumpProblem`s shouldn't make a copy of...

I think this should be fixed by https://github.com/SciML/SciMLBase.jl/pull/72, but I’ll leave this open till that propagates through the ecosystem.

A first fix is in https://github.com/SciML/DiffEqJump.jl/pull/149, see also comments there.

From Slack: we could try to make nomorejumps! more robust though. perhaps by calculating the mean time to the next reaction (which we can do as 1/sum_rate) and if it...

We could perhaps drop the recursive call structure in `Direct` using this. I guess it wouldn’t be good when there are *many* functions to call, but perhaps we can structure...

For most other SSAs we use dependency graphs, so only call a few functions (randomly). I guess in that context this won’t help. But we should definitely investigate for `Direct`.

I definitely brought this up before, though I can't find the issue. It seems like this is now fixed in both solution indexing, i.e. `sol(t, idxs=[sys.x])` and plot indexing, i.e....

You can always register the Julia functions yourself to turn off tracing, but that may mess up applications involving AD unless you define custom derivatives: https://symbolics.juliasymbolics.org/dev/manual/functions/#Registering-Functions-1

Does this handle updating constants in events too?