The donor TTL may run before the recipient has had a chance to clone documents, after starting the migration. That would mean the recipient would not clone any documents.

Therefore, this test, in which we expect recipient-side TTL to be suspended, may not function as intended, because this assertion (that there are documents on the recipient, owing to the recipient TTL being suspended) would fail.

We should implement some synchronization to mitigate this race, or maybe just a check that observes the race and accordingly skips the test, such as the one implemented in the second half of the test.