I had a look at the device tree in Cudy's OpenWrt image and found that they run the SPI flash at much lower speed than upstream OpenWrt (10 MHz vs. 80 MHz). According to the data sheets both flash chips should be capable to work at 80 MHz and beyond, but maybe they got a bad batch of XM25QH128C, or the layout of the SPI lines on the board doesn't work well at the higher speeds, but still better with W25Q128 than XM25QH128C.
Anyway, if you want to give it a try, change spi-max-frequency in target/linux/ramips/dts/mt7621_cudy_wr1300.dts from 80000000 to 10000000, rebuild and flash.
My device with the replaced flash shows the squashfs failures at 80 MHz when heated up, but so far they haven't happened at 10 MHz.
Update: At a SPI clock of 40 MHz I cannot reproduce the squashfs corruption either.