I don't think it reproduces anymore. That was back in times of CEF1 (and in the beginning of CEF3).
Anyways, my "solution" was a combination of
(a) controlled shutdown (signal between host app and libcef via a custom event which simply skipped DCHECK if the shutdown was in progress)
(b) custom exception handler installed while shutdown does its thing (set before CefShutdown and removed after).
(c) dynamic stack analysis on certain Win32 calls (yes, I know...) which decided inside exception handler if the SEH was "expected while we are in shutdown and we're doing this and if that sequence appears etc etc => then ok, we know about this and don't bother user". Not something I was proud of (except was a good exercise on how play god with Win32 internals, keep the user happy and consolidate our crash dumps database).
So - not at all good candidate for a PR, I'm afraid