[alenlpeacock]

That's a number we hit on experimentally. Constraining bandwidth below that, currently, causes the network backup part of the system to perform very, very poorly, resulting in your data not getting out to the system in any sort of reasonable timeframe.

This is in part due to factors that we have some control over and can tweak to allows less capable links to make progress (and we're hard at work on this) and partly due just to plain old physics: if you want to be able to push N GBs of data to the network, with some redundancy factor R, and allowing for self-healing of locations that may leave the network due to failure or otherwise H, and also allow some amount of access from other devices in the cooperative network, in some given timeframe T, there's some math that indicates how much bandwidth is needed theoretically, even with everything optimized.

The good news is, I believe we have some wiggle room here to optimize out still.