AF Learning During Open Loop?

I am running a 2011 sti with all the bolt-ons, Blouch Dom 1.5XTR, and professionally tuned. I've been doing these tuning courses primarily to diagnose/catch mechanical problems from data logs during track days and autocross. I'm a bit perplexed by the most recent data log from an autocross event this weekend; AF Learning 1 applied a -3% fuel correction during open loop, or at least what I think should be open loop. The TPS is 100% and Req Torque is the max of 425 for my setup, however, the car continues to apply a -3% correction. This did not happen at last month's event. The only difference is that I was running a higher gear, so the Load is consistently low at the beginning of these throttle applications (~100 g/s). Today I data logged a few street pulls to try and simulate the same conditions as the car saw at autocross (starting around 3K rpm in third gear) and could not duplicate the AF Learning correction during open loop. The street logs show that AF correction and AF Learning zeroed out as soon as the TPS is 100%.

In the "Open Loop vs Closed Loop" lesson Andre says that long-term trims can affect open loop eventually. However, he says this after a discussion about fuel trims exceeding 5% mean that there is some additional tuning work to do. I understood this to mean that fuel trim will only be applied to open loop if the trims are significantly off (i.g. >5%).

Question 1: If long-term fuel trim is less than 5% (i.g. my 3% situation) would it be normal for them to be applied during open loop? If so, should I be worried that the trims are leaning my AFR out during hard driving?

Question 2: If these long-term trims are being applied in open loop, why wouldn't the ECU apply any fuel trim the next day during highway pulls in the same gear at the same speed and at the same engine load?

Thank you for any insight you can offer.

AF learning can or cannot affect open loop on your particular vehicle, depending on how the tune was set up. That is the tuner's choice.

Where AF Learning break points are for the multiple different learning zones, is also the tuner's choice, so I couldn't guess where yours are set.

The AF Learning break points are based on airflow not calculated load, so that's likely why you didn't see a correlation.