My dissertation utilizes something known as the ICAP Framework developed by Mickie Chi. In essence, Chi breaks down learning activities by their levels of engagement, or Interactive, Constructive, Active, and Passive. For example, watching a recorded lecture without any other activities would be Passive. Higher modalities of engagement will provide higher learning gains, or I > C > A > P.
To that end, Passive activities are simply not enough; they provide some learning gains, but as we're seeing, you need something more. However, some students are not ready for Interactive exercises. For my dissertation, I liken this to the common approach in Intro CS courses where the instructor introduces a new topic, like loops, then expects students to immediately be able to implement it. My philosophy is that we need those Active and Constructive exercises, things like repeating steps (or drilling) which is done in other technical skills; however, when we branch into higher cognitive areas, they're immediately shut down as "not good enough".
I argue instead that we operate within templates, or we behave like we've trained, something stemming my my martial arts background. Right now I think 'virtual learning' has an issue where they're not willing to make this slight upgrade in engagement, for whatever reasons, because its not the magic silver bullet to fix all problems. Rather, in order to improve it, virtual learning needs to add those additional lower level practices into their instruction, otherwise we'll continue this vicious cycle.