I finally went a little crazy and spent several hours working on this problem with multimeters, digital bench power supplies with current meters, fully isolating usb power, monitoring drops and resets...
I have what I now appears to be a solution.
The bottom USB port issues were resolved for my by simply providing adequate grounding to the port in addition to the added positive as suggested.
I bridged the USB V- pins on the USB to T2, and found both ports to become equally stable as a result with low and high current devices.
With that problem out of the way, I then started trying to figure out the issues with the board resetting under conditions like plugging in a wifi dongle, or when wifi dongles hit high tx/rx.
Even with 220uF cap in line (in either direction) and the added ground, the board resets (and sometimes freezes) still intermittently persisted.
In frustration I removed the F1 and F2 polyfuses entirely to ensure the ports were only getting current from the added wire. Much to my surprise, the resets persisted.
The only logical conclusion was that the initial current demands of higher power devices like the USB-N10 were actually trying to draw backwards through the board even at this very early input point, thus robbing the board of power (and/or tripping the main polyfuse).
So! In addition to removing the F1 and F2 polyfuses to isolate the ports, I added a diode to prevent and any backwards drawing from the board.
The board is now rock solid (provided you are feeding in a 1.5A+ power supply). I can boot up with devices in either port, hot swap them over and over etc. I can no longer find any situation that can cause a board reset, as the USB power is completely isolated from the board.
Notes: The diode I added a PR1007 that I stole from a perfectly working power supply in my rage-hackery. It is rated at 1A forward current and prevents up to 1000v reverse voltage which is certainly overkill. I am sure a diode that is easier to obtain and far more suitable for this use case exists. It would also probably be fine to leave the F1 and F2 polyfuses in place with this solution, but I have no real reason to put them back now.