Collector does not catch some exceptions while browser is active, this causes browser procceses to stick around

[simonft]

I think not closing the browser process in the case of an exception, but catching the exception so not letting the lambda die, leaves files in /tmp/, which fills up the space and causes subsequent runs to fail.
See https://themarkup.sentry.io/issues/4012291048/?project=4504775057014784 for an example of the filled up disk space.