Preventing container escape in an interactive Linux learning platform

I'm building a platform where people can learn the Linux command line. My backend is Django, Redis, and DRF. The frontend is React. I'm using xterm.js to pair with the Linux images.

After a bit of research, I've realised that container escape is a real possibility, which can lead to my host being compromised. Here are some of the steps I've taken to prevent this and other forms of potential abuse:

• Read-only root filesystem

• gVisor

• Dropped capabilities

• pids_limit=64

• Resource limits

• Socket proxy

• Zero Docker socket exposure

• Separate lab host

• mem_limit=128

• nano_cpus=250_000_000

• security_opt=["no-new-privileges:true"]

However, I feel this is too restrictive, and even complete beginners will have too little to learn.

What measures can I take to enable root in the images without having to worry too much about container escape? What other security measures can I take?