The ValidateResults (and IndexValidateResults) structs are the primary way results of a validation pass are gathered and communicated to the user.

Unfortunately, due to the way these classes are designed, there is no standard way to communicate that a given finding detected by the pass should be communicated to the user in a certain way, leading to some inconsistencies:

• The "validity" of various objects is not treated in a standardized way, e.g. the validity of index results and results of the total validation are not aligned (SERVER-82278), certain errors are not treated as invalidating the results (SERVER-92361)
• errors reported for index validation are sometimes duplicated in multiple fields of the validation result (SERVER-89857)

We propose to solve this problem by refactoring the ValidationResult class to be opinionated w.r.t. the way that generated errors are reported. Specifically:

• The validity of the result should derive solely from the presence (or absence) of errors
• No duplication should occur in the reporting.
  • If there are any errors in an index, then the main collection validation's list of errors should only note "index XYZ has errors".
• bounds on the size of error lists should be respected for all sources of inconsistencies
• Where possible, the format of errors should be regularized for easy ingestion by automated tooling.