If you find that notifications aren't getting through, but you get no feedback or validation problems, check the rpush logs for EPIPE errors and reconnections. An occasional lapse is to be expected, but if you get one on every request it means that Apple is dropping the connection before the rest of the batch can be sent.

The most likely explanation for the drop is that you're sending a sandbox token to the production service. It could also happen if your payload was too long, but rpush validation would normally catch that first.

Naturally, the dev token is usually first in the list, so nothing comes through and the whole system looks broken.

I don't think it's possible to tell the difference between a sandbox and a production token, so the only way to be sure of avoiding this is to do as apple recommends, and run a separate instance of your provider (and/or token-gatherer) for each push environment.