@zacchiro file a bug to kernel (upstream), emphazing that it is reproducible. Probably a maintainer will get in touch with you, to provide debugging patches. Note: there were also configuration and tools to safe oops into memory (hoping on memory not resetting), but I don’t remember more.