There is a memory leak when running the 'Big Table Browser Memory Leak' test.
The initial version of this test has been modified to consecutively load a page with a large dataTable and then navigate to an empty page, this iteration taking place for 250 times. This test has been automated and can be run in Test Runner using the next steps:
- build the test application war file (ICE-2619.war) using the ant target: 'clean tomcat6-regression'.
- deploy ICE-2619.war to tomcat.
- build the selenium.war file (containing the selenium core files), and deploy to tomcat. For this test the pre-built selenium.war file is attached to this JIRA.
- load in Chrome the URL: http://localhost:8080/selenium/core/TestRunner.html ; then click on the 'Go' button that will load the test case in testRunner.
- open Task Manager and write down the memory usage of all the Chrome processes.
- start the test suite by clicking in testRunner on the 'Run all tests' button.
- at the end of the tests, write down again the memory usage of the Chrome processes, and continue watching the memory levels to see if they keep accumulating.

Here are my results:

1) Chrome19 tests:
Start time: 10:44AM
Chrome processes (total memory usage): 91,000K

End time: 10:52AM
Chrome processes (total memory usage): 787,000K

After the completion of the test there is a decrease of memory usage to ~ 720,600K, then a constant increase of ~4K/sec that occurs as long as the browser is open.

2) IE8 with Chrome Frame tests:
Start time: 11:00AM
Chrome processes:
27,000K
25,000K
IE processes:
22,444K
17,256K

End time: 11:04AM
Chrome processes:
34,404K
723,484K
When the test ends, the Chrome memory usage decreases immediately by ~300K, and then starts increasing again by ~4K/sec; this increase does not appear to stop unless the browser is closed.
IE processes:
23,584K
17,316K
No increase of IE memory usage after the completion of the test.