The Prototype and Script.aculo.us work could end up ensuring compatibility with T5, but nobody has yet investigated Dojo compatibility, so none of this affects T4 compatibility.

Honestly, for T5, I would definitely leave Selenium in. It has a good community and great momentum, warts and all. Once the HtmlUnit JS library compatibility work gets to a more advanced stage (maybe two or three releases down the line?), it would be a good idea to investigate offering both Selenium and HtmlUnit (or one of its derivatives) as options. I won’t go into all the trade-offs involved right now. I’m planning on writing a series of web app test framework comparison articles in the near future, and one of the comparisons will be between HtmlUnit and Selenium.

As far as failure information is concerned, it depends on the kind of failure you’re referring to. If you mean that your assertions fail, then you get whatever information your test framework (JUnit, TestNG, etc) provides. If you mean a bug in HtmlUnit which would not occur in a native browser, most failures are caused by JavaScript host object implementation. In this case HtmlUnit gives you line information on the specific JavaScript that caused the failure, including an integrated JS / Java stack trace.