Not sure if this is a good proposal, but I saw a few cases where some integration test methods in ITestDynamoDBMetadataStore timed out simply. Specially, the one keeps failing me today is testAncestorOverwriteConflict. I increase the timeout to 200s and it works for me happily. Am I using VPN and a slow home network, I'm afraid so.
The time out rule, as inherited from base class HadoopTestBase, is 100s by default. Though that 100s time out default value can be overridden in base class via system property test.default.timeout, that's system wide affecting all other tests. Changing that time out value for one test is no better than overriding in this test ITestDynamoDBMetadataStore explicitly. I think doubling it to 200s would not be very crazy considering we are testing against a remote web service, create and destroy tables sometimes.
[EDIT]: I meant 100s when I said 10s. Sorry for confusion.