Okay, well, I figured it'd probably be a quickie so I watched the video now.
First, diagnostics mode should be engaged when the ECU is off.
I think I've mentioned it before, but if not the MKIII and other TCCS equipped Toyota's are notorious for rich or varying AFR at idle. It's very common. Typically 13 is the richest you'll see and around 15.5 the leanest. Also the ECU uses a richer open loop fuel map after a warm start than it does after a cold start. Temporarily inserting a 4.7K to 10K resistor in place of the ECU temp sensor will ensure that the ECU runs the cold start fuel map. Only do this for diagnostic purposes and only diagnose when the engine is running from the same initial state (ie after restarting once the car is warm).
Don't rely on the car's tach for an accurate RPM measurement. Get a calibrated gauge on it. I've seen the dash unit read hundreds of RPM high and low in the past. IMO, Vf diagnostics are rather worthless as there are many factors which can affect the results and it's really only good for diagnosing gross failures in that specific mode.
You'll find more detailed information here:
http://www.well.com/user/mosk/Vfsignal.htm
Feel free to replace the O2 sensor but until you're POSITIVE the wideband doesn't have a voltage offset, is properly calibrated, set for the correct fuel type and the sensor is good STOP RELYING ON IT... Hell, put tape over it. Once you're sure it is 100% you should seriously consider using it to emulate the narrowband signal to the ECU and ditching the actual narrowband sensor. Narrowband sensors aren't worth much when it comes to diagnosis they're a switch with no grey are or middle ground between results: RICH - Perfect (300mV window right in the middle) - LEAN
From what's shown in the video I'm inclined to believe the ECU isn't even entering "closed loop" operation. I've already detailed the main conditions under which it isn't in "closed loop" (or in other words when it is running "open loop").
There are no error codes present, correct?
Lastly, make sure the cold start injector isn't leaking.