Projection Over Convex Sets (POCS) is one of the most widely used algorithms in geophysical data processing to interpolate seismic data. Whilst usually described as a modification of the Gerchberg-Saxton algorithm, a formal understanding of the underlying objective function and its implication for the associated optimization process is lacking to date in the literature. We show that the POCS algorithm can be interpreted as the application of the Half-Quadratic Splitting (HQS) method to the  norm of the projected sought after data (where the projection can be any orthonormal transformation), constrained on the available traces. Similarly, the popular, apparently heuristic strategy of using a decaying threshold in POCS is revealed to be the result of the continuation strategy that HQS must employ to converge to a solution of the minimizer. Finally, we prove that for off-the-grid receivers, the POCS algorithms must operate in an inner-outer fashion (i.e., an inverse problem is solved at each outer iteration). In light of our new theoretical understanding, we suggest the use of a modern solver such as the Chambolle-Pock Primal-Dual algorithm and show that this can lead to a new POCS-like method with superior interpolation capabilities at nearly the same computational cost of the industry-standard POCS method.