From @zelig:

this test really should not time out, it is either set to too small timeout duration or there's a bug.

src: #1243 (comment)

How to reproduce

1. Create a branch.
2. Modify .travis.yml to run just the Race Detector job, but the whole suite. Timeout does not happen if the test runs just by itself. (Remove everything else from - include group.)
3. Push that branch to ethersphere and make sure your job gets triggered. Travis builds
   => The job should fail. (TestDeliveryFromNodes github.com/ethereum/go-ethereum/swarm/network/stream)

Test run time

If run alone

--- PASS: TestDeliveryFromNodes (77.97s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_2_200_skipCheck_true (4.40s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_2_200_skipCheck_false (4.51s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_4_200_skipCheck_true (5.92s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_4_200_skipCheck_false (5.76s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_8_200_skipCheck_true (8.99s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_8_200_skipCheck_false (9.27s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_16_200_skipCheck_true (19.27s)
    --- PASS: TestDeliveryFromNodes/testDeliveryFromNodes_16_200_skipCheck_false (19.87s)

Questions

• Do we care about the added complexity of if testutil.RaceEnabled?
• Do we expect to learn anything new if we run not just 2 and 4 nodes, but 8 and 16?
• Does that - having more nodes - worth the extra 60s added to total testing time? Even if we can run the tests, should we?
  * Is this investigation and possible fix the best way to use our time? obsolete: tests are passing after cmd/swarm: fix resource leaks in tests ethereum/go-ethereum#19443