CI job on a Windows machine does not execute all steps of the `script` block

Summary

CI/CD job on a Windows machine does not execute all steps of the script block when a gradle command has executed prior to this. When executed on a Gitlab.com such scenario lead to a job timing out without any reason. On a managed GitLab instance with a docker-windows executor such jobs either terminate early without an error but not executing all script steps or terminates with an error that suggests that a corruption of the command by the runner has taken place (see related issue)

Steps to reproduce

Configure GitLab CI/CD repository with a Job primed to be executed on a Windows runner
Ensure that gradle is available on the target machine
Execute the following:

    script:
        - gradle init --type java-library
        - gradle build
        - Write-Output "1"
        - Write-Output "2"
        - Write-Output "3"
        - Write-Output "4"
        - Write-Output "5"
        - Write-Output "6"
        - Write-Output "7"
        - Write-Output "8"
        - Write-Output "9"
        - Write-Output "END"

Example Project

Project on GitLab: https://gitlab.com/azuregos/gradle-runner-bug

Branches:

main (with a reproducible bug)
no-gradle (same pipeline but without a gradle command being used that always works correctly)

Example of the failed pipeline:
https://gitlab.com/azuregos/gradle-runner-bug/-/jobs/2079216393

Example of the pipeline that passed:
https://gitlab.com/azuregos/gradle-runner-bug/-/jobs/2079298938

What is the current bug behavior?

Runner does not execute all steps of the script.

What is the expected correct behavior?

An entire script block must be executed