Additive synthesis time stretching

Hi, any DSP scientists here? I was thinking of an additive resynthesis based time stretching algorithm similar to IL Harmor, and I'm a bit puzzled by how it works FFT gives you magnitudes and phases for each frequency, right? We can just resynthesize those by adjusting magnitudes and phases of each harmonic. But... phase is measured in symmetric units (degrees), and there is no way of knowing if it went forward or back, and using wrap detection is unreliable. What Harmor did is derive frequency from phase based on the difference. It does phasiness, too, but it is not mutable by the user, exists internally and only used for audio resynthesis mode. I'm mainly interested how can we use the phase information to calculate frequency for each harmonic and most importantly how to calculate this weird sub-phase trick Harmor uses. Has anyone thought of this yet?