Struct RenderFailure 

pub struct RenderFailure {
    pub message: String,
}

Inserted the first time Backend::acquire returns AcquireError::Fatal — the render loop has permanently stopped producing frames (a lost device, a destroyed surface, anything the backend itself judged unrecoverable). Once this exists, begin_frame stops calling acquire at all, so the same fatal condition isn’t re-triggered and re-logged every tick forever.

The framework deliberately does not panic or exit on your behalf here — a hard crash isn’t always the right response, and only your application knows whether the right move is an error screen, a full backend re-init, or something else. Check for it explicitly wherever that decision belongs:

fn on_render_death(failure: Res<RenderFailure>) -> Option<()> {
    eprintln!("rendering has permanently stopped: {}", failure.message);
    Some(()) // .once() — react exactly once, not every tick thereafter
}

app.add_system(SystemStage::PostRender, on_render_death.once());

RenderPlugin declares this as provided, so a system with a hard Res<RenderFailure> requirement (as in the example above) waits quietly for it rather than panicking at startup over a resource that, in the common case where rendering never fails, is correctly never going to appear.

Fields

message: String