The ability to accurately detect congestion events in the Internet and reveal their spatial (i.e., where they happen?) and temporal (i.e., how frequently they occur and how long they last?) properties would significantly improve our understanding of how the Internet operates. In this paper we present Pong, a novel measurement tool capable of effectively diagnosing congestion events over short (e.g., ∼100ms or longer) time-scales, and simultaneously locating congested points within a single hop on an end-to-end path at the granularity of a single link. Pong (i) uses queuing delay as indicative of congestion, and (ii) strategically combines end-to-end probes with those targeted to intermediate nodes. Moreover, it (iii) achieves high sampling frequency by sending probes to all intermediate nodes, including uncongested ones, (iv) dramatically improves spatial detection granularity (i.e., from path segments to individual links), by using short-term congestion history, (v) considerably enhances the measurement quality by adjusting the probing methodology (e.g., send 4-, 3-, or 2-packet probes) based on the observed path topology, and (vi) deterministically detects moments of its own inaccuracy. We conduct a large-scale measurement study on over 23,000 Internet paths and present their spatial-temporal properties as inferred by Pong.