Hmm. $0.18 per hour for 4GB of RAM is a rather steep. A c6g.large on AWS would run you less than a half of that, for the same number of CPUs and memory. A quarter if you're happy to use spot instances. Might be neat to have an OSS tool that spins up an EC2 (or Azure or GCP) instance running Code-Server[1] (VS Code running natively on the server, but presenting the UI in the browser), with a given git repo and credentials.
docker run -p 8080:8080 -v "${PWD}:/workspace" mltooling/ml-workspace
Cocalc-docker also includes Git, Jupyter, SSH, a collaborative LaTeX editor, a time slider, but no code-server or VScode out of the box:
https://github.com/sagemathinc/cocalc-docker
docker run --name=cocalc -d -v ~/cocalc:/projects -p 443:443 sagemathinc/cocalc
Admittedly, the storage costs would be higher.
[1]: https://github.com/cdr/code-server