ProvisionerSuite.TestProvisionerStopRetryingIfDying fails intermittently

[xenial] provisioner_test.go:720:
[xenial] stop(c, p)
[xenial] provisioner_test.go:184:
[xenial] c.Assert(worker.Stop(w), jc.ErrorIsNil)
[xenial] ... value *errors.Err = &errors.Err{message:"failed to process updated machines", cause:(*errors.Err)(0xc420beabe0), previous:(*errors.Err)(0xc420beabe0), file:"github.com/juju/juju/worker/provisioner/provisioner_task.go", line:180} ("failed to process updated machines: catacomb 0xc4202a71f0 is dying")
[xenial] ... error stack:
[xenial] github.com/juju/juju/worker/provisioner/provisioner_task.go:923: catacomb 0xc4202a71f0 is dying
[xenial] github.com/juju/juju/worker/provisioner/provisioner_task.go:180: failed to process updated machines
[xenial]

and

[xenial] [LOG] 0:00.902 DEBUG juju.apiserver.provisioner SetInstanceStatus called with: params.EntityStatusArgs{Tag:"machine-1", Status:"allocating", Info:"failed to start instance (container failed to start and was destroyed), retrying in 10s (10 more attempts)", Data:map[string]interface {}(nil)}
[xenial] provisioner_test.go:720:
[xenial] stop(c, p)
[xenial] provisioner_test.go:184:
[xenial] c.Assert(worker.Stop(w), jc.ErrorIsNil)
[xenial] ... value *errors.Err = &errors.Err{message:"failed to process updated machines", cause:(*errors.Err)(0xc420ad9360), previous:(*errors.Err)(0xc420ad9360), file:"github.com/juju/juju/worker/provisioner/provisioner_task.go", line:180} ("failed to process updated machines: catacomb 0xc420692930 is dying")
[xenial] ... error stack:
[xenial] github.com/juju/juju/worker/provisioner/provisioner_task.go:923: catacomb 0xc420692930 is dying
[xenial] github.com/juju/juju/worker/provisioner/provisioner_task.go:180: failed to process updated machines
[xenial]

The traceback doesn't make any sense. The string "stop(c, p)" does not exist in provision_test.go

The string *does* exist in Juju 2.2, but *not* in develop, and both of those patches are targeting develop.
Do we have an issue with unclean builds?

Should be fixed by https://github.com/juju/juju/pull/8007/commits/203d949196ba67da6c596ad61382afd79c4df073

Ok, so it seems that most likely Andrew's patch which removes the stop() calls, just landed in between Anastasia's PR that failed the suite, and her second submission (which did actually land). Thus the failing one was still merging develop, just a version that still had the stop() calls.

Andrew's patch at the very least makes this traceback no longer valid. So we might as well consider it fixed.