Occasionally (maybe once a year for the last few years) a bad event gets passed to our snowplow connector. This breaks Fivetran's processing of all other snowplow events. It's usually just one event out of the 10 of millions you all process for us on a monthly basis. Someone on your side then has to go in and manually delete the event. On average this has taken ~2 weeks to resolve. In the meantime we aren't getting any new snowplow events. There's not any data loss, but the timely delivery of this data is critical for downstream consumers like our internal lead scoring system and A/B tests on our marketing site.

We'd like to see the snowplow connector improved, for instance by dumping bad events into a separate table, or your internal process for removing bad events improved, so that there's minimal downtime from bad events. Thanks