can someone explain to me how brickwalling happens at lower levels
By "lower levels" I assume you mean lower recording levels, not lower sound levels (e.g., a quiet show).
From my new-found understanding, the menu's GAIN setting controls a first stage, and the rocker-switch LEVEL setting controls a second stage. So for a typical loud show you may find yourself setting GAIN=HIGH, and then reducing the LEVEL setting to prevent the PEAK light from blinking. But what's going on is you're overloading the first stage (controlled by the menu's GAIN setting), invoking its limiter feature to "brickwall" the first-stage output. Once that's happening, you are misled into thinking you got a good combination of settings. It seems that the first-stage limiter produces an output level that then requires a LEVEL setting at ~70 to avoid the PEAK light from coming on. So the guideline is to not use a LEVEL setting below ~70, if instead you can reduce the GAIN setting, because if you find a LEVEL less than ~70 is necessary (with GAIN higher than LOW), it might mean that you're overloading the input stage, and your recording will be brickwalled.
My take-home (using the internal mics - YMMV) is to start with GAIN=LOW, and LEVEL at a high setting (I started with 85 last night), and after the first song (or during the sound check), back down on the LEVEL as necessary. If you're stealthing, and can't get to the unit to reduce the LEVEL after the first song, set DUAL at -12dB and hope for the best, expecting to use that lower-gain recording for production if you guessed too high on the LEVEL setting. (At least, if you record 24bit/96kHz, and intend to publish 16bit/44.1kHz, you have numerical room to change the gain in post on the resulting DUAL recording without too much loss of fidelity.)