I replaced the MK5 R32 exhaust to the cats with one out of a MK6 golf R , which is straight bolt on job, but has no resonator.
Now I get small constant KR on one cylinder only. It's only -1.5 , but I know it is not real because when I retard the timing  by 5 degrees it still want to retard by 1.5 every time.
It only occurs when the muffler valve opens, which I set to 4500 RPM. I assume some pulses are interfering with the detection.

What would be the usual way a tuner gets around this problem?  Adjust RKRMX2N (this cylinder is the only one mapped to this) or adjust KE4N?

 I understand that RKRMX2N is only used when rkr > RKRMX2N * 2ˆ(vkr-6), so I need to log these to make sure, otherwise it is pointless to alter the map.

Quick update, Im still investigating. Not so sure it is the exhaust but it is not knock.
The retard shown is not triggered by B_kl which surprised me me.
Adaptive kr is off, and  wkra  array are all zero when I dumped RAM the end of the log.
Wkr suddenly gets a value  of -1.5 and when it advances by 0.75 (when the counter hits zero), it is immediately brought back down to -1.5. Again, not by B_kl. I patched in some small custom code to log the B_kl by incrementing a counter, at free RAM location, as it is too fleeting for normal logging. Looking at IDA atm,alas very slowly.

nice work

Optional Leading Cylinder (LZ)

The leading cylinder function is enabled:

- On exceeding a cylinder-specific speed threshold KRNLZ, above which the cylinder has poor knock detection, this cylinder is led by the cylinder with a good knock detection........

If the code word CWKRLZFK = 1, the retard for the led cylinder is determined according to the following minimum selection:
i is current(led) cylinder, j is leading cylinder, WKRLZOF is the pesky -1.5 that kept rearing it ugly head.
wkr_i = MIN (wkr_i,

wkr_j) + WKRLZOF_i