pub fn seed_watermark_to_tip(db: &Db, checkpoint: u64) -> Result<()>Expand description
Seed the broadcast pipeline’s resume watermark to checkpoint when
it has none yet, so it starts at the live tip rather than genesis.
The broadcast pipeline is registered like any other (it persists a
watermark and resumes from it), but it is not covered by the formal
snapshot or floor_unrestored_pipelines,
so on the first run after a restore it would have no watermark and
the framework would start it — and, since the shared ingestion start
is the minimum across pipelines, all pipelines — back at genesis.
Seeding it to the current tip avoids that. A no-op once a watermark
exists (subsequent runs resume normally), so callers pass the tip
and skip the call on a fresh database with none.