As discussed on the IRC channel with @dcramer yesterday, it seems that in some instances, stack traces can be missing from some of the events in my Sentry instance. The trace instead shows a single frame, which is a raise directive (in Django), which should obviously preserve the stack trace.
If this is something that you'd consider to be out of scope of Sentry, and caused at a higher level (either by Django or Python, at a guess), I'd be happy to take it up with those guys, but the responses from there previously has always been "it can't be us, take it up with Sentry" so it would good to have something like this ticket to point to.
More about the environment:
The stack traces we're seeing instead of the actual error:
TypeError: must be string, not datetime.date
Stacktrace (most recent call last):
File "django/core/handlers/base.py", line 121, in get_response
The raw JSON data for this particular event
As mentioned on IRC, this is not limited to a single error, but seems to happen for approx. half of the exceptions we see in production. It seems to consistently happen for the same error cases, however.
I'm going to close this, as after some further investigation I don't think this is an issue with Sentry (or Django, or Python). The culprit here seems to be either gevent or psycopg2, and it appears to be a manifestation benoitc/gunicorn#214.