After observing the conditions when this test fails, it appears that
some (but not all) signals are available. As these signals are generated by a
rule via a bulk create, the odds of us retrieving signals in the middle
of that bulk creation is very slim (but not impossible).
The crux of the error here was: we wait for signals to be generated, but
not the ones that we need. Specifically, we are waiting for a single
signal to be available, but since we are asserting on sequences of
signals, we need several to be available to us.
While not perfect (because the signals we receive are not technically
guaranteed to be sequence signals), increasing the number of signals
that we wait for before proceeding should be sufficient to prevent this
failure state. In debugging, it was observed that every test returning
9-10 signals succeeded, while it was possible for the test to return
only one signal and fail.