LMS6002 Manufacturing Issues

We have a bunch of designs using the LMS6002 device. It’s great on most of them but on a couple we have manufacturing yield problems. On those boards the LIME does not respond to SPI accesses reliably so it cannot be setup properly. In these cases the voltages, clocks and resets are all as they should be.

We have triple checked the footprints and masks against the recommended and gerber eval board files and they match,

On the bad boards we’ve XRAY’d the devices and they look excellent, with very little voiding.

One thing we’ve spotted is that on some bad ones the XRES12 signal is not stable. On a good board it yields a steady square wave output.

Has anyone else had manufacturing issues with this package?