Analysis of what is causing models to break.

[SamuelBrand1]

No description provided.

[SamuelBrand1]

At the moment as discussed elsewhere and f2f with @seabbs the working hypothesis is that wild priors are causing numerical instability.

As of current state of #445 we can run and gather evidence of this over all the scenarios and model configs.

[SamuelBrand1]

I think this is worth doing as part of SI explaining how we arrived at specific priors for specific models/scenarios.

IMO, in good practice papers with Bayesian analysis the selection of priors is an important topic. The added complexity here is that we are looking a comparison between a fairly large number of models, scenarios and time horizons (100s) and therefore we should document prior selection with that in mind.

[SamuelBrand1]

I'm lowering the scope to identifying the problem.