This is likely my issue since users and developers generally recommend TNR after AOF.
However, after upgrade to v8.0, my gpt graphs for Sentinel-1 SLC data no longer performed TNR, and did not provide any notice to the failure: No errors or messages different than v7.0.
v7.0 correctly processed TNR regardless of the order of these operators.
The new graph simply has AOF prior to TNR and now correctly performs TNR in both v7.0 and v8.0.
I suspect this may be the same for GRDH processing. But I have not tested.