The idea is lots and lots of low orbit satellites with cell-phone-like handoffs as they zip across the horizon. LEO is only 200 miles high or so, which is ~1ms for light to travel if they can make it work.

Existing satellite internet goes out to geosynchronous orbit which is much, much farther away.

I believe they are planning on an orbital altitude of between 700-800miles. I'm not a network engineer so take this all with a heavy helping of salt but I'd wager that the overwhelming majority of networked application that are dependent on extremely low latency are also relatively local(intracontinental). So given this assumption is accurate the increase in latency should be minimal for these types of applications.

I've also read that the plan is for the constellation of satellites to number at 4425 satellites strong and for them to be using frequencies between 10GHz and 35GHz.

FWIW SpaceX claims the networks latency to be around 35ms.

Source: https://arstechnica.com/information-technology/2016/11/space...

Current satellite internet is provided by a small number of very far away satellites (geostationary). Musks plan is to use a large number of leo satellites that are much closer (and cheaper to launch and replace).