> It probably thinks the id check is going to be better to limit the result
> set.
>
> How many records are there for id > 1935759 ?
About 40 million or so.
> vs
> How many records for company_id = 4 and source_model_name =
> 'CommissionedVisit' ?
>
> If this is a common query you could probably do a multi-column index on all
> 3 columns (id, company_id, source_model_name) - but if company_id and
> source_model_name have a low number of distinct values, then it's not going
> to help.
Both of those will indeed have a pretty low number of distinct values.
Looks like I need to figure out something else.