Origam.Architect.Server’s Workbench.DoModelChecks runs ModelRules.GetErrors over the whole model in a background task on every start - and then throws the result away.
Nothing reads errorFragments, nothing reads the persistenceProvider local; there is no return value, no field, no log, no event.
It came in with the HTML Architect (#3120) as a copy of the desktop ArchitectWorkbench version, where the parts that consume the result (the confirmation dialog and FindRulesPad) were commented out. Later cleanups removed the commented code and left only the computation.
It is not a safety net for a broken model either: the background task’s error handler only logs, so nothing about the startup fails because of it. And the model is read from disk anyway for the reference index, so what is actually wasted is a full parallel model-rules pass over every schema item on every start.
I suggest removing it - the rule validation itself is worth exposing on demand instead.
Btw, I am currently working on an “on-demand” model validator, so I will fix this along the way.