Because we separate loadtest results per URL now, we can move the error
tests into the regular mixed bag of loadtests - we will be able to tell
from the misspelled URLs when we hit a regression in that area.
We should be able to do similar things for JWT tests, but we'll need
more infrastructure here.