In Code Insights missing timeseries values are not recorded. That is to say for a repository set {A, B, C}, the matches {A: 5, B: 2} will save data for A and B but not for C. In the case of tracking migrations this hurts the experience somewhat, because values can never become zero (missing values are not recorded). To make it slightly more confusing, just-in-time insights (live preview) will show zeroes.

As an example (bottom is JIT and top is after saving):

We don't want (and can't in many cases) to fill out the missing value set with zeroes. For large insights this would be especially difficult / wasteful. Here are some potential ideas to improve this:

1. For scoped insights populate the missing values and save with zeroes.
2. Track the set of queried times (without repo cardinality) and join this set to the result to coalesce 0 in place of nulls