- [ ] Update check script:
  - [x] shellcheck on shell scripts
  - [x] `go vet` on go files
  - [x] `golangci-lint` on go files
  - [x] Run `go fmt` on all .go files
  - [ ] Eventually, run unit tests

- [x] Reimplement fixup.sh
  - [x] run shfmt on all shell scripts (and the files they `source`)
  - [x] Run `go fmt` on all .go files
  - [x] Run ~~prettier~~ biome on all .ts files and maybe others


- [x] Adapt master program so that it reads configuration from command line
  args instead of from environment variables
  - Should have sane defaults
  - Adding new arguments should be easy and obvious

- [x] Add wrapper script to run master program (so that various assumptions related
  to relative paths are safer)

- [ ] move `master-bin` into a subdir like `master/cmd` or whatever is
      idiomatic for golang programs; adapt `master` wrapper shell script
      accordingly

- [ ] Add unit tests all over the place.
  - ⚠️ Huge task - needs breakdown before starting

- [ ] flesh out the `sync.sh` script
  - [ ] update framework-managed node
  - [ ] update framework-managed pnpm
  - [ ] update pnpm-managed deps
  - [ ] rebuild golang programs

- [ ] If the number of workers is large, then there is a long lapse between
      when you change a file and when the server responds
  - One solution: start and stop workers serially: stop one, restart it with new
    code; repeat
  - Slow start them: only start a few at first