Weird 'run-sync!' behavior

[qtrfeast]

If I have multiple 'run-sync!'s in my test, the second one doesn't seem to update state until after a small amount of time. The following is just an example, but if I need to put together a sample app that duplicates this weirdness I can.

If I have:

(deftest test-thing
  (let [app (app/build dataflow-description)
            app-model (render/consume-app-model app (constantly nil))]
       (app/begin app)
       (testing "app model test"

         (is (test/run-sync! app [{msg/type :set-value msg/topic [:some :path] :value "thing"}
                                  {msg/type :set-value msg/topic [:some :other :path] :value "blah"}]))

         (is (= (q '[:find ?v
                     :where
                     [?n :t/path [:affected :path]]
                     [?n :t/value ?v]]
                   @app-model)
                [["maybe some derived thing"]])))

       (testing "look in data model"
           (is (test/run-sync! app [{msg/type :set-value msg/topic [:some :path] :value nil}]))
           (is (nil? (-> app :state deref :data-model :some :path))))))

The last is assertion in will fail.

If I eliminate the first testing form (and therefore remove the first call to run-sync!), the last is will pass!

Even weirder, if I simply add a (Thread/sleep 100), leaving the first testing form, before the last is, it will pass!

Any ideas?