It's not clear to me that you when you'd want the same exceptions to apply and when you wouldn't, or how you'd indicate that, so I'm not sure what the right way to do this would be. 

In practice, just from looking at the history of the _exceptions file, it seems like you'd always want the same exceptions.

That could be true, but I feel like we don't really have enough data points to be sure or to give me any guidance in what the design might look like. 

You've been spending more time than I looking at this, though, so if you think you've got a sense of what might make sense, feel free to propose something / code it up / whatever to give us something concrete to talk about.

How about adding a new flag to the exceptions which indicate whether they should apply not only to the test suite of that name, but also to those where that's the 'test'? The default could be either true or false, but I think the option should be there to toggle it.

For the particular examples of browser_tests, viz_browser_tests, network_service_browser_tests, and site_per_process_browser_tests, we do *not* want to run all four test suites everywhere exactly the same way. 

The three most obvious reasons for this are that (a) the command lines might vary, and the exceptions are what we use to capture that variance, (b) often these suites are either still being brought up, or only make sense on some platforms, and (c) we may not have the capacity to actually run all four of them on particular configs, and we get enough coverage in the CQ by only running them on some of the configs.

So, I am still kinda skeptical. It could be that a flag is the right thing to do, but like I said, I'd want to see concrete proposals to evaluate.