Treat the long audio recording as the backbone. For every camera start, create a visible or audible sync event if possible. A clap is fine, but on an action shoot you can also use a horn, a helmet tap, a brake-light event or anything that creates a distinct waveform/visual cue.

For longer clips, check another matching event near the end. If the offset is still the same, you have a simple start offset. If it has changed, you are dealing with clock drift.

Do not confuse frame-rate interpretation with true device drift either. If one camera was converted or interpreted at the wrong frame rate, the sync error can grow even if the hardware clocks were fine.

I would sync the untouched originals first. Only after you know the timing relationship should you decide whether any clip needs a proxy or intermediate format.

In VSDC I would put the continuous audio on the main timeline, then align each camera segment to a matching transient. Once a clip is aligned, group or otherwise keep its offset stable before doing detailed cuts.

If you also have GPS, a sharp acceleration or stop can become a second independent synchronization reference. That is useful when the external recorder has weak ambient sound and the camera mic is mostly wind noise.

File organization matters a lot here. I name folders by device, not by “good clips” and “bad clips,” because duplicate filenames from two cameras are common. Then I keep a small text note with the start order and any known sync event.

That sounds tedious until you have CAM1 removed link and CAM2 removed link in the same project.

Using the recorder as the continuous spine is the part I was missing. I was trying to make every camera clip the reference, which gets messy after each restart.

I will also check a second event near the end of the longest clip before assuming one offset is enough.