Like i mentioned with KR leveks over about 5.5 i see the same trend you show. But with KR levels below 5 the ignition timing will not follow a trend relating to it. Look at your KR levels. It well above 5 for most of that log.
If that were the case, it would seem logical once knock retard drops below 5, that spark advance would return to a pre-knock retard level. In my log, it doesn't. Are you implying, by chance, that it takes an initial KR jump over ~5-6 for spark advance to initially be affected, and from there spark advance follows knock retard?
Once I settle into cruise, KR is reporting 5.3 and spark advance is 29. KR then drops to 2.8 and spark advance has grown to 32. A delta of 2.5 and 3 respectively; not exact for sure. It is hard to do an exact comparison even when cruising because the spark advance will change by 1 degree here and there for seemingly no reason. KR gets as low as 2.1, but only for perhaps a single data point. The spark advance is 33 and when knock retard jumps back up to 5.6, the spark advance drops to 29. A delta of 3.5 and 4.
Also, to include another data point... here a sub-picture from a later part in the log...
I didn't include it, but RPM is stable varying about 30 RPM across this portion of the log. Engine load starts at 22% at the beginning of this portion and settles at ~18%.
So, in this particular log, the initial KR value is a paltry 1.1 but the spark advance drops a full 11. Now, this may however be an artifact that the load drops from 21% to 18% in conjunction with spark advance. However, that is counter-intuitive because at a constant RPM, a drop in load should generally lead to an increase in spark advance. Also, BAT drops 17F across this picture which may also be a factor.
That aside, KR drops to 0.4 and spark advance is at 27. KR then jumps to 2.5 and spark advance drops to 25. A delta of 2.1 and 2 respectively. With each decay and jump of KR, there seems to be a corresponding change in spark advance. What is odd about this log... is that at the end, engine load drops by only 1% to 17%, RPM is still holding constant, and despite the KR reading having now spiked to >5, spark advance increase to 33. This flow with the idea of less load more advance, but such a small drop in load doesn't seem to correlate to such a relatively significant increase in spark advance.
I'll certainly do some more logging this afternoon... and post back with more logs. Just FYI, I am using the calculated load PID -- I may skip logging BAT and log both calculated and absolute load as calculated load may not reflect well given the scale we are on.