We’re seeing a similar problem with 420 “Exceeded connection limit for user” errors here. I posted a thread explaining the details a few days ago over at https://dev.twitter.com/discussions/27717
I’ve now tested that this is happening both with/without gzip Accept-Encoding headers.
We’re using Twython as our client library, which uses Python Requests under the hood. So it doesn’t look like we’re having any problem with the tweets being split across HTTP chunks. We don’t see any issues with JSON decoding the data we receive - just intermittent 420s on reconnections.
@mattklein123 - any suggestions as to whether the 420 problem is related to the change in HTTP chunking, or is some other unrelated change in the way stream.twitter.com is operating? Any help would be greatly appreciated as all our attempts to figure out from here what’s changed have failed so far.
Thanks!