So you can probably fit three to five one-bounce reflections in your field of view. Then there’s the “infinite” number of higher-order reflections, but given that the probability for a photon to be absorbed rather than reflected is maybe 5% every time it hits a mirror, the additional reflections don’t really contribute that much beyond maybe 20 bounces. So I guess the odds could improve to something like 1/10,000 as a really rough guesstimate.