Nils, any change to ace:ajax or ace:panel that could account for this change in behaviour?

This is an expected change from the reduction of the default execution scope of ace:panel behaviors from @all to @this.

As has been previously noted, using getter incremented values to determine if a region is executed or just rendered is a inadvisable technique prone to several sorts of false negatives and unclear results. I thought they had been removed from our tests? Apparently not all.

They don't serve a strong purpose either, they only serve to show that an execute & render has been large enough to include the given components or not.

More reliably, the server side should just inspect the request parameters for the execute and render scopes and test there for the desired broadness. This can the result can be rendered to the page for evaluation by selenium (or in the case where there is execute but no render, send result via JavascriptRunner).

This doesn't test to the same degree as the existing implementation, it assumes the given scopes are executed as indicated, but that is the frameworks responsibility anyhow. What we gain is clarity of expectations and results, and assurance that we won't get false positives / negatives from coincidental getter access.

FYI: A JIRA to establish a more robust test technique for ace:ajax @execute tests has been created: ICE-9054.