In the future, it is a good idea to log only the parameters necessary to diagnose the issue. Logging excess parameters only serves to cut down logging resolution. /PSA
Took a look at your logs. The second log is only cruise, so not a very useful log in my book if only to look at the LTFTs and STFTs. Both fuel trims look reasonable at cruise (unlikely you have a gross vac leak). Your DAM (IAM) drops immediately after ramping up the load during throttle on. Firstly, no forum warrior truly knows the way Subaru ECM strategy splits FBKC, FLKC, and IAM. The fact that IAM pulls timing without registering FBKC and FLKC may be due to the load cells in which the knock is present. Since you are at low load, FBKC and FLKC are disabled. I pulled open a stock 2005 STi map to demonstrate:
Admittedly, I don't know what the load cells Cobb's map is set to register FBKC/FLKC, but I would assume it's similar to stock. Excessive knock sensor chatter at low loads is usually due to rattles in the engine bay, not necessarily detonation. A useful parameter to log in the future would be the Knock Sum parameter. I don't know if this parameter is available in pre-CAN 32 bit ECUs (it wasn't available in the 16 bit ECUs). So, a couple thoughts:
1. I would check your engine bay (particularly the heat shields) for rattle that may contribute to false knock
2. Next log should include Knock Sum if available
3. When was the last time the Spark Plugs were replaced? Bad spark can contribute to a non-repeatable flame front as can carbon buildup in the valves in pistons. Have you ever sea-foamed the car?
4. Without a WBO2 sensor, it's difficult to diagnose this, but throttle-on knock is often caused by poor tip-in tuning. Tip-in changes based on injectors and modification to the intake system. Can you confirm that you are not running an aftermarket intake, post-MAF hose, or turbo inlet?