I've researched this a bit over the years, and also did the ADR analysis as a young boffin on a Hawk which the RAF lost following a practice turnback.
It's clearly not a desirable thing to do, but it is noticeable that of all the people who have flown turnbacks, virtually none have ever been killed. On the other hand, virtually none ever got to use the same aeroplane again either!
To make it work, you clearly want to start right at the beginning end (not far end) of the runway. It's not going to work at-all unless you have a fair bit of height available and good visibility to see exactly where you've come from and where you're going.
After that, generally the approach that works is to turn one way first using maximum available roll rate, then fairly rapidly the other way (creating a "teardrop", turning right on the buffet all the way, at a reasonable speed and bank angle.
In still air, from an average runway, my experience (from experiments at a safe height) is that a typical light aircraft, with a very sharp pilot, might just get away with it from around 600ft, a slow microlight about half that, a fast jet 1-3 times that depending upon speed and stall margin at take-off. A good stiff headwind helps.
And with all of this knowledge, would I brief and attempt it? Almost certainly not - I can think of very few scenarios where I'd actually attempt the take-off, and landing ahead isn't a better option. Put another way, if landing ahead isn't a viable option, why am I attempting the take-off?
I understand that this manoeuvre is still taught to, and regularly practiced by UAS QFIs but primarily, I suspect, as an excuse to give them interesting flying and keep their handling skills sharp, than as a realistic emergency drill.
Finally echoing what PilotDAR said - plan it, brief it, execute it. Any pilot failing to brief appropriate emergency actions before take-off, or to stick to his brief in an emergency, is one I'd not care to be sharing an aeroplane with.
G