For our internal builds, our build process is roughly the following:

1. build the product and pack (this builds packages)
2. install dev certs
3. run non-helix tests
4. install sdk+workload for testing (this builds a subset of packages, again)
5. run helix tests
6. run Pack, sign and publish (this builds all packages, again)
7. build workloads

This is not good given we are re-building packages in step 6 after tests have already run against the previously built packages (from steps 1 and 4) which potentially invalidates the test run if something is slightly different in that step 6. I found this when investigating #5396 as it seems like this is what is making all of our tests pass, but the actual final packages we push have a bug. We should fix this by:

• Ensuring that packages in our build are only built once (to avoid duplication)
• Ensure that after running tests, nothing is ever re-built to ensure the integrity of the packages and not invalidate testing.

cc: @radical