Originally Posted by
The_Doctor
Only if the music is precisely metronomic. Musicians rarely are, there is subtle rubato (changes around some fixed tempo) and actual a tempo changes too, in almost all recordings not synced to a click track. Jet Black of the Stranglers is a 'metronomic' drummer when he wants to be, but even he does it. It takes only a few seconds worth of parallel metronome to prove it.
One thing that can screw up markers on saving so they're wrong on reloading is time based marking. If you can set the time gauge to samples and not beats/ticks, or real times, the save/load operations will preserve the markers. This is true for Sound Forge, and likely for anything where the beat/tick resolution is a lot cruder than single PCM sample duration.
If a VBR MP3 is loaded for audio, and markers are being screwed up even when no seeking is being done randomly (or otherwise) in the VBR file, then check that the MP3 data isn't being recompressed back to MP3 on saving! I doubt that Beyond or pretty much anything will do this, unless it's an audio editor that is assuming you want it to do it. Check anyway though because it's bad for the sound, worse every time it happens, as well as messing with relative timings.
Whatever the cause of timing error is, try to find out what it is. It may not always be obvious. Evidently not as there seem to be at least two unsolved instances mentioned in this thread already. The one thing to remember is that if the audio SOUNDS right, it is right, because that's the whole point of a working audio subsystem, to make it sound right. Whatever the frame size or rate or compression method, timing should remain accurate relative to start (and to any other marker that is also correct relative to start). This is obvious which is why it's a good reality check. If no seeking is being done, and if audio and video start together, and are clocked by accurate timers, they'll stay in tight step for days. So if these conditions are met, and your timing still gets wrecked, look elsewhere for the answer.