Apparently it was hourly and Inmarsat calculated two arcs (north and south) which it could have taken, but then focus seems to have moved to the southern one in last few days.

I wonder if they flew test flights along both arcs and compared the data.

They didn't need to fly test flights. That would be big $$ and take time.

But apparently, their hardware is installed on plenty of airplanes, so they just needed to study normal flights north and south in the region to establish a baseline to which to compare MH370's data.

And it was that data that lead them to conclude that the southern route was likely the one MH360 actually took.

> I wonder if they flew test flights along both arcs and compared the data.

No, but they compared the present data with recorded data from other flights along the same routes for comparison.

Well, there wouldn't be flights along the Southern route, exactly, since it led nowhere.

So I'm guessing they extrapolated from routes that were in that general direction, like Sri Lanka - Perth.

It's likely there's a lot of assumptions at work in their calculations, but it might be good enough.

Some possibilities http://www.gcmap.com/mapui?P=jnb-per http://www.gcmap.com/mapui?P=per-rio

I should have been more clear -- they compared routes and aircraft velocities that could be expected to produce the same radial velocity and Doppler effects, given the satellite's geostationary position.