unable to handle render process crash

[jesg]

steps:

1. run a test
2. crash all chrome render processes via pkill -f "chrome.*type=renderer"
3. try to run another test

expected: next test runs successfully
actual: selenium fails with unable to connect error

notes:

• the render process occasionally crashes in real tests
• the parent process does not crash when the render subprocess crashes

[jesg]

chromedriver is not complaint with the w3c spec on error handling. hence i'd have to parse the body of every response, get the status in the body, and crash the browser if it is one of the fatal statuses.

w3c would make this a lot easier because the bad failures are all 500 errors.

as long as you crash everything after each test suite it should be ok for now because the browser and agent just hang there. this is my current model when running in docker.