Ask HN: How to interpret PyPI download numbers?

For Python packages, 100k per month is clearly better than 0. In the middle of the long tail (1k to 5k per month), what's driving these downloads -- organic users or machine mirroring? Is there a monthly download number below which you red flag a package for adoption?