I'm not sure what general principles would result in our popularity-based heuristics excluding religious beliefs, but including popular scientific theories which we have not evaluated for ourselves. What do you guys think?

The strength of others' beliefs as evidence depends on what you know about how they arrived at those beliefs. If you know that scientists have a general process for establishing accepted truths which involves repeated testing with attempts to falsify their hypotheses and find alternative explanations, then you can take established consensus as evidence proportional to your trust in that process. Likewise, if you know that people tend to come to religious consensuses due to early indoctrination and community reinforcement, you should take religious consensuses as evidence proportional to your confidence that those processes will tend to produce true beliefs.