Determining flow boiling critical heat flux (CHF) using mechanistic models or empirical correlations requires careful validation with the aid of reliable databases. But, while many new databases are being made available in the literature, the methods used to detect CHF vary greatly, producing different CHF estimates for the same fluid and operating conditions. The variations in detection method are the result of both heated wall design and criteria used to terminate an experiment in response to wall temperature excursions. This study investigates the interfacial phenomena preceding the occurrence of CHF for flow boiling with a finite inlet vapor void. Experiments are conducted with FC-72 in a rectangular channel that is heated along one side. Temporal records of the heated wall temperatures are used to track the complex transient response of the heated wall, and identify differences between temperature excursions associated with momentary localized dryout and those with true CHF. It is shown that the flow enters the channel fully separated, with a liquid layer sheathing all four channel walls surrounding a central vapor core. At high heat fluxes, a wavy vapor layer begins to form beneath the liquid layer adjacent to the heated wall, and cooling is provided mostly through wetting fronts associated with the wave troughs in accordance with the Interfacial Lift-off Model. However, depending on mass velocity, inlet quality and flow orientation, conditions may arise that cause breakup of the heated wall liquid layer into ligaments that are entrained in the vapor core. This phenomenon causes localized dryout and wall temperature excursions at heat fluxes well below CHF, but the wall is able to recover from these excursions by a combination of reattachment of ligaments with the heated wall and lateral heat conduction within the wall itself. Recommendations are made concerning construction of the heated wall and CHF detection in pursuit of reliable CHF data. (C) 2013 Elsevier Ltd. All rights reserved.