Testing is a continuous process it is need to be examined frequently that once code worked it keeps on working. this is why sanity tests and regression tests are the most impotent tests in the ALM process. Needless to say that every new feature test written should eventually be added to regression tests.
The automated test result are the best tool we have to decide “Go” or “No Go”. When I worked at Zend Technologies I got each morning a report saying which test passed and which failed but the I had the report also highlight the tests that are newly failed and those were brought up at the Stand-Up meeting. Those highlighted failures were then combined to the code changes done the day or hours before.
The tests which fails regularly and have low importance were brought again to the planning in order to assign time to get them fix.