My recommendation would be to look at the machine driver and see if its making additional calls to the DSP driver that is not made in other machine drivers such as the bdw-rt5677 (Samus.) That might indicate an additional code path that might be getting exercised in your context that isn't used in samus which is causing your problems. If you find something you can always copy it over to samus to see if it causes the same breakage. So yes definitely look. Usually the suspend/resume paths aren't that long, but I would search the whole machine driver for anything that can alter state.
The only significant difference I see in the machine drivers is that the clock dividers are smaller in this bdw-rt5660 case, the bitclock is 4.8 MHz v. 2.4 MHz in the other bdw-rt5677. It's not very clear to me why there is a need to have different clock and SSP settings, but since the patch that was suggested to be reverted touches the clock generation maybe there's a connection?