I thought it had got better, but also I've just adjusted to work around it. Here are some timings today, Julia 1.1, cold start to first plot:

    $ julia -e '@time (using GR; plot(rand(20)))'
      3.931433 seconds (10.38 M allocations: 521.292 MiB, 6.44% gc time)
    $ julia -e '@time (using Plots; plot(rand(20)))'
     19.498644 seconds (57.26 M allocations: 2.844 GiB, 8.07% gc time)

Running with less compilation:

    $ julia --compile=min -e '@time (using GR; plot(rand(20)))'
      0.375836 seconds (368.83 k allocations: 20.190 MiB, 1.65% gc time)
    $ julia --compile=min -e '@time (using Plots; plot(rand(20)))'
      4.302867 seconds (6.41 M allocations: 371.485 MiB, 5.07% gc time)

But, as you say, it's much much quicker once started, like 1-5ms per plot.