Network error can result in apparently successful runs with missing data

[jbothma]

It looks like a network error doesn't necessarily mean a run is aborted. That is really bad. If I don't check that all the data is there manually, I might upload data with a missing piece.

datapackage-pipelines==2.0.0
Python 3.7.1
Arch; 4.14.84-1-lts

I ran the pipeline unchanged two times.
after run 1 the data for one of the sources was missing in the concatenated output.
After run 1 I deleted .cache and ran again.
Run 2 was fine. All the data was there

I think cache is disabled, whatever is the default.