...the only thing that seems to work is to feed the signal into the noninverting input, plus add a 1M resistor which is connected to my Vdd/2 rail, and remove small 10p cap to ground at the input. The C3/R3 II C7R7 filters are extremely important, to the overall tone..C7 is .15u so I had to wrap a couple caps in paralell to get close to that value. Its definitely in the ballpark of the sound I've heard on vids of the MG10..got that Marshall growl.

Using a single supply and the circuit as shown in post #1 try a simple bias on the TL072 like this. Input and output coupling caps are required to block the dc voltages.
SG

 

I don't know why you have two input coupling capacitors and why you have two RC networks in the feedback to ground because one input coupling capacitor and one RC network feedback to ground will perform and sound exactly the same. My simulation shows them:

 

The point was to show how to bias the TL072 without changing the circuit as is. Yes you can eliminate R2 and the .47uf cap.
SG

 

read this guys..its the technique i used for avoid inginstability issues. amp works fine http://www.analog.com/media/en/technical-documentation/application-notes/AN-581.pdf