Bad Scipy/Powell H fits when geometry is not yet downloaded after IPH upgrades #84
Description
Fitting H with Scipy/Powell is fairly bad before geometry has been downloaded: 
After geometry comes down, the fit goes back to normal:
No other changes were made between running of these cells.
The problem does not seem to occur (or is at least nowhere near as pronounced) when fitting with Dynesty. Below is another pre-geometry fit, first with Scipy/Powell:
The same integration with Dynesty:
If I had to guess I would blame the hoops we have jumped through to get uncertainties.