Bayes CHIME generating unbelievable output may be connected to unrealistic posteriors

[Mdraugelis]

LGH recent simluations show a long growth of Hosp census through July

While all other entities show a decline.
For example

MCP	HUP

This could believable if BayeChime showed a worse beta and social distancing values for LGH compared to MCP and HUP. However, this is not the case.

LGH	HUP

Instead, LGH posterior distributions are distinctly different for "Days from infection to recovery" and "Days from exposure to infectiousness"

LGH	HUP

From a distinct GOF perspective, this makes sense since the growth at LGH is long and slow. This might indicate a long period of the infectious period. But it doesn't make sense that these two values would be different across regions.

The likely short term fix is to hold the days of "Days from infection to recovery" and "Days from exposure to infectiousness" to a narrow distribution for all regions.