Google reported a spike in update request errors (pver4) starting on January 13, which corresponds to when we enabled V4 updates on Nightly. According to them, they are receiving protocol buffers that can’t be parsed by their frontend. Chrome had this problem when they weren’t using the right base64 encoding. This seems to corresponds to about 4% of update requests: https://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2017-01-19&keys=google!other!mozilla!__none__&max_channel_version=nightly%252F53&measure=URLCLASSIFIER_UPDATE_REMOTE_STATUS2&min_channel_version=null&product=Firefox&sanitize=1&sort_keys=submissions&start_date=2016-12-21&table=0&trim=1&use_submission_date=0
Padding should be the root cause at the first glance because we  use the same padding policy as chromium .  https://dxr.mozilla.org/mozilla-central/rev/3cedab21a7e65e6a1c4c2294ecfb5502575a46e3/toolkit/components/url-classifier/nsUrlClassifierUtils.cpp#354  https://chromium.googlesource.com/chromium/src.git/+/3db3676fdfc65bfd15436db3e86f5d5c1bc4536a/components/safe_browsing_db/v4_update_protocol_manager.cc#226
(In reply to Henry Chang [:henry][:hchang] from comment #1) > Padding should be the root cause at the first glance because we  use the ^ NOT (sorry for the typo) > same padding policy as chromium . > >  > https://dxr.mozilla.org/mozilla-central/rev/ > 3cedab21a7e65e6a1c4c2294ecfb5502575a46e3/toolkit/components/url-classifier/ > nsUrlClassifierUtils.cpp#354 >  > https://chromium.googlesource.com/chromium/src.git/+/ > 3db3676fdfc65bfd15436db3e86f5d5c1bc4536a/components/safe_browsing_db/ > v4_update_protocol_manager.cc#226
According to Google, the protobufs we are sending are actually fine. So, now that bug 1332770 has landed, we should wait a few days and see if we still see the same unusual rate of errors from V4 requests. If we don't, we can close this bug. Thanks for looking into this Henry. As always, your help is most appreciated.
This issue should have the same root cause as Bug 1332780 that some of the errors may because of connection errors, we could verify this after Bug 1332780 land. I'll mark this as duplicate as Bug 1332780