It has been proposed that the Kosi River continuously migrated >113 km westward across the surface of the megafan over the last two centuries. Examination of a number of old maps published between 1760 and 1960 shows that during most of this period the Kosi River occupied a position slightly east of the megafan axis. The apparent channel movement shown in these maps is oscillating in nature and not unidirectional. Instead of encountering deposits left behind by a sweeping braided Kosi-like stream, a preliminary study of the uppermost 2–5 m succession in the north-central part of the megafan reveals overwhelming dominance of meandering stream deposits. Assuming the existing notion of Kosi River migration, the rate of deposition averaged over ∼100–200 years for the uppermost ∼5–10 m of megafan deposits, works out to be unusually high (>50 mm/y). All these observations question the soundness of the hypothesis of rapid westward migration of the Kosi River over the last two centuries. The existing facies model for the uppermost 8–10 m of the megafan deposits also appears untenable. The three-dimensional geometry of the Kosi megafan is similar to those of typical alluvial fans, but with much gentler gradient (0.05°–0.01°) and with larger area (>10,000 km2). Based mainly on the patterns of paleo and modern channels recognised in the satellite images, three major accretionary lobes can be identified on the Kosi megafan. Relative age of the lobes determined from the truncating relationship of the paleochannels indicates a random shift of the trunk channel forming these lobes. Similar multilobate form and evidences of random switching of the loci of lobe aggradation are also found to be common in the Tista and Taquari megafans. The factors known to favour avulsion and the results of the recent simulation studies of alluvial deposits are inconsistent with the notion of unidirectional shift of the channels for more than 100 km across the entire megafan surface. This study suggests that the relocation of the Kosi River in the past was through random nodal avulsion rather than systematic unidirectional shift. The recent avulsion of the Kosi channel by a large distance to the east follows this expected pattern. Further study and age dating is required for a comprehensive understanding of the depositional dynamics of the megafan and the pattern of channel movement on it. Future flood predictions and disaster management plans should be based on such comprehensive understanding.