I think the user impact of this will be that constraint checks using SPARQL (“subject type”, “value type”, “distinct values”) won’t always work, and users will not be shown some violations of those constraints even when they should be; but if I’m reading DelegatingConstraintChecker::getCheckResultsFor() correctly, these errors won’t abort the whole constraint check, and other constraint violations should still be shown if I’m not mistaken.

I'm wondering if more requests are coming because one or more impactful constraints have been added. Anyone got a hunch? Worth doing the detective work?

@Lydia_Pintscher we're waiting on you to tell us how important / urgent this is.

To me it looks like SPARQL type checks generally went back to normal around January 4th (Grafana permalink):

In T325730#8509750, @Gehel wrote:

@Lydia_Pintscher we're waiting on you to tell us how important / urgent this is.

Generally a part of the constraint checks not working is bad for Wikidata because editors don't get shown notifications about issues in the Item they are looking at.
But as Lucas said things seems to be looking ok again.

@dcausse I think we can close this if the metrics look good from your side too? (I don’t know what I’m looking for in the Envoy Telemetry Grafana dashboard.)