I think the right path forward is to not try to gain agent-level performance between pipeline runs by allowing them to share the same workspace over and over.
In order to avoid full re-cloning every time between the various runs of a single pipeline on an agent, the agent should have a cache directory where git repositories can be stored, and then perhaps hardlink or clone from the local disk storage in the pipeline workspace.
Picked @markewaite's brain about this, and reference repositories look like a good way to solve this problem. I'll have to figure out the right strategy to use here however.
Identified as part of #33
I think the right path forward is to not try to gain agent-level performance between pipeline runs by allowing them to share the same workspace over and over.
In order to avoid full re-cloning every time between the various runs of a single pipeline on an agent, the agent should have a cache directory where git repositories can be stored, and then perhaps hardlink or clone from the local disk storage in the pipeline workspace.