Struct TailRequest 

pub struct TailRequest<'a> {
    pub store: Arc<dyn ObjectStore>,
    pub store_prefix: &'a str,
    pub volume_root: &'a Path,
    pub subjects: &'a [String],
    pub tier: Tier,
    pub owner: &'a str,
    pub page_size: usize,
    pub rpo_target: Option<Duration>,
}

Everything start needs. Mirrors HydrateRequest field for field where the meaning is the same, because they are read from the same declaration.

Fields

store: Arc<dyn ObjectStore>

The object store both the claim and the data live in.

store_prefix: &'a str

Key prefix for this workload. The claim sits here; each subject hangs one level below — the identical layout crate::hydrate restores from, and not a parallel one, so a tail and a hydrate of the same workload cannot disagree about where the bytes are.

volume_root: &'a Path

Host directory the named volume is bound from.

subjects: &'a [String]

Volume-relative database paths, in declaration order.

tier: Tier

owner: &'a str

Label recorded in the claim. Diagnostic; see ClaimRecord::owner.

page_size: usize

Page size of the source databases. Required rather than sniffed for the reason StreamConfig::page_size is: a wrong guess silently produces frames nothing can replay.

rpo_target: Option<Duration>

RPO the caller’s scheduler promises, carried into the watermark so a missed round is visible as a breach rather than as silence.