
Just in case the CS pin on the XMOS device and related IP is making this pin behave as open drain (OD). Ran into this issue with Cypress last week for an undocumented port pin. That is, the external pull-up was mandatory to see a logic high.
Confirm again that the power rails are stable. Perhaps the XU208 consumes more current than the original XHRA processor?
Silabs clock generator (PLL) is ok for at least the required 24M line?
CPU swap is clean and no shorts for the replacement? If practical, consider to review with an X-RAY system.
The ADUM1085 is an open-drain reset supervisor. Does the output of this supervisor go high ok? Would not hurt to apply an external pull-up if you do not have in for this circuit. Believe that the XMOS device has an internal pull-up for the RST_N pin but again, will not hurt to apply one to be sure the CPU is released from reset.
XU208 devices sourced from an authorized supplier? Digikey / Future Electronics are ones that we know of are ok. To be sure you are dealing with remarked empty packages.
Perhaps you have already tested this but if you remove the external flash device off your PCB - then do you see the CS pin of the XMOS (SPI master) / flash device toggle?
also
The XU208 features a metal belly pad which could be fun to solder. Confident that this required metal belly is soldered? This is a ground pin to the device. This missing pad could be a root cause of the non-working boards.
Any chance you have bonded out the test points (TP) to match the original XHRA PCB? If yes, then you could wire up to an external XTAG3 tool to further debug if the XU208 CPU is alive and functional on the custom PCB. See pins 60..63.
