Draft: Change decoding checks for corrupt data
With previous fixes to the decoding, we observed a divergent number of errors with increasing mu. This MR makes (hopefully) the checks a bit more readable, and fixes the linearity issue.
The number of errors is expected to be more-than-linear, in the sense that bit-flipping only happens if certain links send clusters (linear behaviour) and there are clusters in links before that link so that the error gets picked up.
Illustration:
Original (link, channel) sent: (0, 30) (12, 30).
Bit flip 12 -> 4
New (link, channel) sent: (0,30) (4, 30)
No problem seen
Original (link, channel) sent: (0, 30) (5, 28) (12, 30).
Bit flip 12 -> 4
New (link, channel) sent: (0,30) (5, 28) (4, 30)
Problem!