Currently the RPC test framework does not include CL engine api calls. This means the test chain we spin up in tests don't actually produce canonical blocks and progress the l2 head. The flashblocks state processor can "pretend" a canonical block has come in, as we currently do, but this limits the full scenarios we can run and prevents writing certain types of tests that are becoming necessary to write.
For a better end to end test setup, we should have a proper e2e setup where the chain also can produce real canonical blocks and progress the L2 head. It's okay if these need to be manually triggered rather than producing a new canonical block at a fixed interval.