Rather than merging Observations when returning from functions, which is very nice and functional, we should be more pragmatic, strive for simpler code, and use the Runtime to save observations.
The overall objective here is to cleanup and review dslx, then start using it in production and rewrite experiments using it. In turn, by doing that, we'll enable collecting better throttling metrics for experiments.
Rather than merging Observations when returning from functions, which is very nice and functional, we should be more pragmatic, strive for simpler code, and use the Runtime to save observations.
The overall objective here is to cleanup and review dslx, then start using it in production and rewrite experiments using it. In turn, by doing that, we'll enable collecting better throttling metrics for experiments.