direct to pdf, https://dl.acm.org/doi/pdf/10.1145/3733699

The paper title is "Sandboxing: Foolproof Boundaries vs. Unbounded Foolishness"

An abstract ending "or does it?" is not a good abstract. The conclusions should be summarised as well as the questions.

Are you worried only about mistakes, or also about fraud? (And if the former, are you sure?)

That calls for different strengths of sandbox.