The management of the dsh group files used to select hosts that should be synced by scap has been moved to a resource collection based system in operations/puppet (https://gerrit.wikimedia.org/r/#/c/179121/). This may lead to hosts being in the mediawiki-installation group which are known to be broken.
To prevent error fatigue on the part of deployers, it would be nice to find a reasonable way to let the deployer know that although an error was seen syncing host X it was in some way expected. @Joe has suggested that checking icinga may give a good indication of expected down/broken hosts. This would require some amount of reworking the error signalling mechanisms internally in scap.