'Homebound' children are unable to attend school for illness-related reasons. To lessen their predicament, schools have begun experimenting with 'telepresence robots' that can enable remote participation. While promising, we know little about the use of telepresence robots in practice. To begin to redress this, we draw on 159 semi-structured interviews to explore the experiences of 37 child users of the robot 'AV1' in Norwegian schools. The children's experiences varied, with some benefitting greatly and others not getting any benefit from using the robot. To explain these variations, we reconstruct the robot's "critical component structure" -- that is, the assembly of sociomaterial elements that determines whether and how the robot works in practice. We also explore the benefits of using the robot when these critical components align. In so doing, we provide in-depth knowledge about the potential and prerequisites of using telepresence robots in schools -- to the benefit of users, producers and scholars of telepresence technology.