During integration of snowplow JS tracker we found out that with slow connection we got canceled status with error : net_error = -3 (ERR_ABORTED) for requests with > 5sec time.
I’d be tempted to first debug why the requests are taking so long to get to the collector, e.g., is the DNS lookup taking too long or is the TTFB (time to first byte) longer than expected?
The request/response from a Snowplow collector should be able to comfortably make a roundtrip within 1 second so a request taking longer than that suggests something is amiss.