Three (combinable) options here:
- We stop bothering with general benchmarks (limited use, do we ever look at / act on the wildly varying results in the CI) and just focus on benchmarking our own code changes locally when it is relevant (and putting the results in the PR body).
- Don't worry about testing at the
Node level at all.
- Create a better benchmarking framework such that we can easily benchmark
Nodes
Three (combinable) options here:
Nodelevel at all.Nodes