Need to investigate whether we can safely intercept and handle openat (and the related *at family of I/O calls). We would need to do so without breaking the atomicity of the calls, which will be a ...