Free-running oscillators mimic the behavior of most vintage analog synths. The oscillators are continuously running whether they are making a sound or not, so when you do trigger notes, each note will start at a slightly different point in the oscillator's phase (the location in the waveform's peaks and valleys at which playback starts). This contributes in a small way to the "imperfect" nature of analog synths that make them sound so unique. That being the case, why would you choose not to run in a free-running mode?
The main reason is that those variations in waveform phase can cause the attack portion of the sound to vary in a way that might not be conducive to the type of sound you're trying to create. A good example of this is synthesized drum sounds. Generally with these types of sounds, you want a consistent sound on the attack. Putting your oscillators into a non-free-running mode allows you to achieve that consistency because the waveform plays back from the same point in the oscillator's phase every time you press a key. This mode is also perfect if you want more precise, modern bass sounds.