Back in 2009 when I was ran my first marathon and then started thinking I could ramp up my speed, I ran into all kinds of AT problems. Eventually to solve the problem I learned to keep a shorter stride, no over-striding, and a midfoot strike. To do this and not slow down you need a higher cadence.
Since I'm a data nerd, to help with that I bought the little cadence sensor that works with the Garmin(s). Here's a plot of some of my cadences vs pace for 90 days of running back in 2009:
Here's the same kind of plot, but from 90 days prior to now:
My 9 minute cadence has gone from 82.9 to 88.3 steps /min, a 6.1% increase in cadence. The slope has gone from 1.65 cad per min/mile to 2.06. This means for each 2.0 step/min increase in cadence, I pick up a minute per mile in pace.
Essentially, I have a faster cadence at any pace, and I use cadence improvements to get more of my speed when I increase my pace that I did before.
A lot of speed does come from stride lengthening but keeping the cadence up i think reduces the chances for injury. In addition, having an efficient long stride AND the ability to a fast stride gives you more ways extract all the energy out of your muscles in a marathon.
I think a faster stride also increases running economy (the energy cost of running). Running with a slow, bounding stride means you are asking your leg muscles to store the elastic energy of rebound in your muscles for a long time.
Muscles are wet sloppy bio-goop, not metal. They don't store energy well...if you want to get out the energy you put in you need to do it quickly before is dissipates as heat. Hence the evolution of drills like 100-ups and short fast strides that harvest the rebound energy as snappy as possible. Elite runners are in the air all the time and have very very short ground contact times!
YMMV, but I feel if you want to avoid overstriding or other biomechanical problems a cadence sensor to give you feedback and to keep you honest might help.