Hamstring tears seem to have a pretty high rate of recurrence don't they? Your not alone, in fact, I am currently helping out a friend (I am a student of Nutrition and Exercise Science) who is doing his PHD with his research centred on hamstring tears. His hypothesis is that there are actually permanent neural inhibitions which contribute to the negative trends in the force-length curve.
The force length curve is a line graph of the relationship between the force the muscle is capable of producing at various lengths, and research indicates that after hamstring tear, the curve shifts to the left, showing that the muscle is weaker at longer lengths. This seems to be the key to effective rehabilitation and prevention for future re-injuries. Obviously, the study is still underway but here's what the current research tends to show.
First of all, flexibility training does not reduce risks (contrary to popular belief). It seems that eccentrically biased training can actually produce the best benefits to prevent hamstring tears and heres why:
Hamstring tears will almost always occur during the eccentric contraction. In running, this occurs when the knee is near in full extension, as the hamstrings must contract to decelerate and reduce stress on the ACL. So directly, eccentric strength will provide huge benefits to reducing ACL injuries and perhaps even tears themselves, but it goes on.
Eccentrically biased training has been shown to produce in-series hypertrophy, which basically means that sarcomeres grow in line with each other, rather than next to each other. The benefit of this is greater contractility, and a higher rate of muscle shortening. These combined tend to produce a rightward shift in the force length curve, meaning we could potentially restore the negative effects of the tear itself by becoming stronger at longer lengths (where you are at most risk). This indicates more of a muscle architecture related solution, however, neural contributions could also contribute.
Neural contribution could theoretically re-teach us to be active at longer muscle lengths. While this represents a new mechanism for the potential benefits, the protocol remains the same, eccentrically biased training. My friend doing the PHD thinks that it is likely that both neural adaptions and muscle architecture changes are connected, so it makes sense if they are able to be solved simultaneously. Like I said the study is still going on, but I suppose details are not important (of how we would work out whether architecture or neural mechanisms are important, but if you want to know I could briefly explain from what understand).
So practically, the best exercise would be a nordic hamstring curl (google it, but basically a glut-ham raise where you lower yourself eccentrically and push yourself back up to eliminate the concentric portion). A study on Norwegian footballers showed that this practice reduce rates of hamstring tear significantly. There is also evidence to implicate fatigue, so it may be beneficial to remain cardiovascularly fit (well, you should anyway).
As for setting it up, I would imagine 3 sets of around 5 reps (with say a 6 second negative, or just as slow as you can) would be around the optimal range. To high a rep range would probably make you not do the exercise correctly. Ideally, find yourself an exercise physiologist or rehab expert who knows there stuff.
Sure, I am currently involved in a hamstring study which gives me access to the guy doing th PHD and his supervisor, (my lecturer, who by the way, taught me everything I am teaching you), but the use of eccentrically biased training is not THAT cutting edge in the sense that the evidence has been well and truly out there for a while, so make sure the guys know his stuff. Most of all, just be careful dude, don't hurt yourself. Stay safe man I hope this can at least get you going.