Pretty accurate as long as someone isnt loading a "multisport" activity. There is a known issue with Strava separating out the data.

You're going to have to give a bit more detail on what you're trying to discern. Do you mean in terms of the precise location of one person relative to another?

Personally I've noticed that the positioning of people riding in a group on Strava Flyby should be taken with a massive pinch of salt. For example, I've ridden long stretches towing a whole bunch of people behind and looking back on Flyby it'll have me solidly in the middle of the pack, some people a few seconds up the road, some people a few seconds back.I put this down to small discrepencies in the time synced to each GPS device. Sure you can tell we're all riding along in bunch and if/when someone falls off the back/flies off the front, but in terms of exact positioning, there's certainly a significant margin of error there.

I don't know whether that's significant in what you're trying to look at.

I've looked at the flyby's after some races and noticed up to 1 hour time discrepancies. I think it's from the multi-sport issue.

I use it to determine if a break away was as successful as I thought during group rides and analyze what works and what doesn't.

For that, pretty good. If you want to see if people were pack drafting, it might not be useful if the gaps between people are constant. If people are moving past or falling back, it's better.

Otherwise, just the nature of the beast may make people look closer together than they are.

Yeah, this is basically what I was trying to say above.

In terms of relative positioning it's not too bad, but in terms of absolute positioning it would be totally unreliable for that kind of thing.