One possibility is to let the CQ inspect the Tricium results of the last patch set and then block on results from certain analyzers. This integration would be mainly on the CQ side, with needed data available in Gerrit or via Progress RPCs to Tricium. The Progress RPC is currently designed around a run ID, but this could be expanded.

An slightly different approach is to mark results as blocking. This would allow an analyzer to produce both blocking and non-blocking results, but it would also be more invasive in that changes to that separation would require a change to the analyzer.

Note, we had discussed merging CQ and Tricium, making tricium analysis jobs a type of tryjob, and making it so that tryjobs can add comments to CLs.

Then, one way of addressing this would be to make it so that such tryjobs produce comments and are run as part of presubmit in CQ.