HMC7044 failures

We have board we have made based on the HMC7044.  We have had occasional success getting clocks and communications, but for the most part, this is not the case.  We have removed, replaced, tested, and x-rayed the PCB and HMC7044 and the most noticeable thing that stands out is that when measuring the voltage on LDOBYP4 (all others are ok), this appears to float around.  We have changed out all of the LDO bypass caps and in one instance, the device did work for a few days before failing.  The connections are based (and extremely similiar) to the eval board).  Can anyone comment on this situation?