Zeros in unscented covariance
The unscented filter produces zeros in the covariance matrix when including propagators parameters. This isn't the case with the kalman filter whereas I would expect both to behave very similarly.
This is the original forum discussion thread.