Hacker News Reader

How are you isolating filesystem state between concurrent sandbox instances without incurring heavy snapshot/restore I/O penalties?

We use gvisor and cloud hypervisor depending on the use case, which manages these things pretty well by default.