Two main things here:

1. For the tasks, the prior convention was to use the same task prefix for any tasks that shared the same project. So, if the project is acme_insurance, then the tasks would be acme_insurance_xxx. This breaks that, both in calling the tasks things like free_text_hqls_001, and by having different suffixes (hqls, hqhs, etc). Ultimately, if there's a good reason for that, great, but we should probably do that intentionally if we do.
2. There are already some support for fuzzy matching in the harness. It should support things like mismatched column ordering and row ordering, and there's support for approximate equality in numeric fields. (https://github.com/dbt-labs/ade-bench/blob/main/docs/CONTRIBUTING.md#approximate-equality-tests). I'm not sure if this will cover what y'all need, but it might be good to unify those so that it's all handled through a single mechanic.