At the point of committing to developing a custom testing platform there are many options available, and for a researcher without a background in technical systems it may be difficult to work out the best way to implement their testing.
The possibility of a documented near-turnkey solution as per issue #8 would reduce the tendency to go down overcomplicated routes, but it might be worth commenting on them as a short learning pathway, particularly noting:
Keeping long-lived VMs up may require regular maintenance at an OS level
Kubernetes-based solutions may have a very significant initial learning investment that doesn't give long-term returns for someone whose primary work and skillset is science, not sysadmin
Not containerising research software may be a false economy, as once the software has been containerised it's probably in a good position to be used on any one of many testing platforms
At the point of committing to developing a custom testing platform there are many options available, and for a researcher without a background in technical systems it may be difficult to work out the best way to implement their testing.
The possibility of a documented near-turnkey solution as per issue #8 would reduce the tendency to go down overcomplicated routes, but it might be worth commenting on them as a short learning pathway, particularly noting: