The massive dev chart recommendation (9 mins) will be fine. I agitate for the first 60 seconds then 10 seconds a minute, all by inversion except for LF film (which tends to come loose in the tank and damage itself). I'm sure agitating by stirring if you have tanks which allow that is just as good: I just don't do it.

It's unsurprising that your previous results were not good: pushing Tri-X two stops is going to result in hard, grainy negs which will inevitably lose highlight & shadow detail: that's what pushed B/W film looks like, especially pushed Tri-X. It then looks from your description as if you took a picture in difficult lighting conditions (in my house today there's about 4 stops difference between the window embrasure and the wall next to it, and far more for windows lit by direct sun), probably using an averaging meter which is going to be just overwhelmed in light like that.

Ther's nothing wrong with what the massive dev chart suggests: you simply got the results you should have expected. No magic is going to make Tri-X pushed to 1600 not lose detail in cases like that.

Might I also suggest that just using Tri-X at its rated speed is probably a good place to start? Perhaps you've already processed lots of such films, in which case I apologise for suggesting something you've already done. (I've exposed probably thousands of rolls of Tri-X: I don't think I've ever used it at other than rated speed.)