I solved the noise problem for now. I removed the BA 3515/580kv and replaced it with a spare BA 2826/690kv. It is now very quiet in flight.
The rear collet on the 5mm shaft was a bit loose on the 3515. This was allowing the can on the motor to move fore/aft a fair amount. That was the likely the culprit for the unpleasant grinding noises in the turns. I will disassemble the motor to see if there are any scuffing marks on the magnets or stator and I'll possibly replace the bearings while I'm at it. Compared directly, the new 2826 has almost zero fore/aft free-play. I do like that 2826 uses 3 bearings. Seems like 3 bearings would support everything slightly better in a rear-mount, 2-blade setup like mine. There is a lot of cantilevered forces happening.
To satisfy my curiosity, I pulled the Castle ESC data log for my two test flights with the 2826 motor. The log shows this motor only pulled 442 watts and 21.2 amps at peak with the 12x6e APC at 9250'ish rpm. The data indicated that I used around 1380mAh of the (P26a) 6S-2600 li-ion battery, (which was only charged up to 4.15v per cell.) I was relieved to see that voltage sag is not much of a concern with this combo, as the voltage was only pulled down to a few low blips of 3.3v per cell during the overhead-8 and clover. It is reassuring to know that using the 690KV motor, there is plenty battery voltage headroom on tap, and I am nowhere near the 2.7v sag prevention cutoff setting that many discuss. When I checked the battery a few minutes after landing, I saw ~3.6v per cell on my meter.
Glad to find (but not surprised) that the BA 2826/690kv is also a pretty happy combo in this plane. The 2826 tugs this 64oz plane along with just as much authority as the 3515 it replaced. 5.3 lap, using 69ft .018" lines. I'll just fly the Sultan in this configuration for a while and use the repaired 3515 in something else.