Sometimes TG instances are unavailable due to incoherencies in the ontology. These often bypass the ontology's normal CI jobs, because the TG import chain is larger or out of sync. E.g. this happened with OBA for a while (usually fixed by remaking the imports/ directory)
Couple of solutions (not necessarily mutually exclusive):
The ontology takes responsibility for setting up an ontology-specific CI job that tests with an import chain that corresponds to the TG one
We run one giant central TG integration test (e.g. daily CI job)
The disadvantage of 1 is that the config is in two places. The disadvantage of 2 is that it seems rather monolithic and centralizing. But presumably as all the configurations are there already it's easier. TBD
Sometimes TG instances are unavailable due to incoherencies in the ontology. These often bypass the ontology's normal CI jobs, because the TG import chain is larger or out of sync. E.g. this happened with OBA for a while (usually fixed by remaking the imports/ directory)
Couple of solutions (not necessarily mutually exclusive):
The disadvantage of 1 is that the config is in two places. The disadvantage of 2 is that it seems rather monolithic and centralizing. But presumably as all the configurations are there already it's easier. TBD