The behavior of the sidebands during the last lock is quite strange: sometimes they drop to a very low level and stay down showing a certain instability; sometime they are back to a normal and stable level. This trend is common to all the PDs demodulated at 112 MHz (fig 1), and is seen also by B1p, which contains the sidebands for a certain amount (fig 2). There is nothing similar in the carrier, for example B4 (fig 3). At a first look, I cannot find anything in the alignment signals which could explain this behavior.