Our SharedWorker implementation could benefit from tests that explicitly create situations where the RemoteWorkerParent experiences an unexpected ActorDestroy due to process loss. Bug 1514733 is an example of a bug that would have been caught by such coverage. Additionally, IndexedDB and LocalStorage NextGen have also demonstrated that we had poor handling of these scenarios because we had insufficient test coverage.
In that bug I proposed the following test which was too large for the bug, but I think is coverage we want. From https://phabricator.services.mozilla.com/D26963#792396:
I think it's feasible to write a browser test for this where we create a SharedWorker in process A from a page in process A, then join that SharedWorker from process B, then force a crash/abort of process A. This puts the RemoteWorkerController in a bad state. Then we can have another page join the SharedWorker and we want to verify that the resulting SharedWorker is both responsive to a ping and that the browser didn't crash.