Hello,
I have discovered this also. These appear to be two separate issues, the 180 degree phase jump is to do with the PLLs for the RF mixer stages and the small phase shift is due to some random delay. If you follow this link: Application for using the 2 RX channels simultaneously it provides examples of some software which is used on one of the SDR platforms to detect and correct this problem by testing the phase difference using internal RF loopback and correcting it by repeating resetting the problematic block. The method is rather crude but seems to work most of the time.
