What techniques are programmers using to write maintainable integration/functional/acceptance tests?

Integration/Functional/Acceptance tests are inherently difficult to maintain for a myriad of reasons: they take a long time to run, are hard to keep green consistently, can fail seemingly 开发者_高级运维randomly, are cumbersome to debug, etc. etc. However, they can be priceless, if one manages to have a reliable non-unit test suite.

Is having such a suite, merely a pipe-dream? How are teams able to manage a consistently green, low maintenance non-unit test suite?

This won't answer the question fully, but here goes. For Java, I use Junit4 (testing) and EclEmma (code coverage). For Javascript, I use Jasmine (testing) and JSCoverage (code coverage).