The high plasma temperatures expected at reactor conditions in magnetic confinement fusion toroidal devices suggest that near-marginal operation could be a reality in future devices and reactors. By near-marginal it is meant that the plasma profiles might wander around the local critical thresholds for the onset of instabilities. Self-organized criticality (SOC) was suggested in the mid 1990s as a more proper paradigm to describe the dynamics of tokamak plasma transport in near-marginal conditions. It advocated that, near marginality, the evolution of mean profiles and fluctuations should be considered simultaneously, in contrast to the more common view of a large separation of scales existing between them. Otherwise, intrinsic features of near-marginal transport would be missed, that are of importance to understand the properties of energy confinement. In the intervening 20 years, the relevance of the idea of SOC for near-marginal transport in fusion plasmas has transitioned from an initial excessive hype to the much more realistic standing of today, which we will attempt to examine critically in this review paper. First, the main theoretical ideas behind SOC will be described. Secondly, how they might relate to the dynamics of near-marginal transport in real magnetically confined plasmas will be discussed. Next, we will review what has been learnt about SOC from various numerical studies and what it has meant for the way in which we do numerical simulation of fusion plasmas today. Then, we will discuss the experimental evidence available from the several experiments that have looked for SOC dynamics in fusion plasmas. Finally, we will conclude by identifying the various problems that still remain open to investigation in this area. Special attention will be given to the discussion of frequent misconceptions and ongoing controversies. The review also contains a description of ongoing efforts that seek effective transport models better suited than traditional equations to capture SOC dynamics. Most of these models, based on the use of fractional transport equations and related concepts, could prove useful both in reactor operation and experiment control and design.