I think we, i.e. those who already use base 10 and neglecting the possibility of us getting invaded in some way, will continue to use base 10 unless our civilisations fall apart. When we handed off the basics of our system to machines so that the low level tasks could be done quickly, the need for a more efficient system decreased. Out to a ridiculously large number, the basic components of our system are, when coupled with machines, essentially instantaneous.
What's 395847593874382754238754987 * 389756987476347629845 ?
1.5428437e+47
What could any system of numbers give me that would make that operation faster than typing it in?
My old math teacher's objection to this was:
"But what if you don't have a calculator?"
And I was not smart enough at the time to realise there'd be more serious problems in a world where I didn't have a calculator and wanted to multiply large numbers (hey, I was only six.)
But what she ought to have said, when I was objecting to learning the tables, was that it makes sense to be reasonably fast with the basics of a system, so that you can do things beyond linear algebra quickly. If you need to stop and work out the very basics of a system every time you do algebra you're probably not going to get very far. The low level tools you have available influence what you can build on top of them.
However, the faster you are at the low level the less the practical gains are. The value of reducing an operation that takes a minute to one that takes seconds is likely to be enormous - but the value of reducing that second to a half second is not likely to be as significant.
It's similar to handwriting. We had a superior system of handwriting that we used to teach: Shorthand. It was more efficient even than most people's typing. But the additional value of that speed over typing was not sufficient for it to remain.
Unless there's an argument that, say, our being a half second or so faster at the low level will allow us access to some new high-level concepts, I think we're likely to stick with it.