It is pretty obviously not the case that eBPF means shared-kernel containers are comparably as secure as VMs; there have been recent Linux kernel LPEs that no system call scrubbing BPF code would have caught, without specifically knowing about the bug first.