I wrote:
> [ assorted possible workarounds ]
Actually, an easy fix might be to create a 2-column index on
usagestats(type, tid). I think the planner should be able to
use that to produce sorted output for the mergejoin, and you'd
get the best of both worlds, because the indexscan will stop
immediately when it's exhausted the rows with type = 'vehicle'.
regards, tom lane