If the migration hasn't committed when the state doc is read here, the TenantMigrationAccessBlocker check can fail (i.e. time out due to assert.soon) since the migration could continue past the state the check expects. To workaround this, we can either pause the migration or disable replication on the initial sync node so that the TenantMigrationAccessBlocker cannot transition to the next state while the check is being done. Note that we had to do something similar in the startup recovery tests (we used failpoint PrimaryOnlyServiceSkipRebuildingInstances to make the migration not resume after restart).