Where it'll increase productivity or not is at larger scales. If Servo comes out at production quality in some reasonable time (call it 2 years or less from now, given where they are), able to function on the general web most of the time, and it's also faster than current renderers by a significant margin, and with some small fraction of the development effort (probably even if you include the entire Rust development effort; do not underestimate the effort poured into the existing engines!), I will consider Rust's story proved. We've now had many people take cracks at writing renderers, in the fastest language of the time for the task (C++), and I daresay they've converged at fairly similar levels of performance with little reason to believe the renderers are going to get much faster. Said renderers have had massive optimizations applied to them, and this Servo that I'm expecting to be much faster will actually be version 1.0, basically.
If Rust is to work at all, there will very, very rapidly be a class of development that it will be actually insane to write in anything but Rust... but that will only be the very high-and-large end. (It will take many more years for this fact to penetrate through to the development community at large.)