Fleet ## Parent

[Jhazy33]

Parent

#1 (fleet spec)

What to build

A wire-level cap on catch-up replay: an optional per-session event limit carried beside the hello cursors (absent = full replay, so old daemons behave exactly as today), with the exact shape as agreed with upstream in KenKaiii#2 (fallback if upstream is silent: the cursor-side limit). The daemon honours the cap when replaying from its session log and signals truncation so the app can indicate older history exists. Demoable: a reconnect after a long absence replays at most the cap per session on a new daemon, and full history on an old daemon.

Acceptance criteria

• New app + new daemon: replay per session is capped; truncation is signalled to the app
• Old daemon + new app: full replay, unchanged behaviour (cap field absent/ignored)
• Old app + new daemon: unchanged behaviour (cap absent)
• Cap shaping is a pure function over cursors with table-driven tests
• Implemented per upstream's answer in Design: multi-machine pairing (fleet sidebar, per-machine status, rename, per-machine push) KenKaiii/pew2#2 (fallback: cursor-side limit)

Blocked by

• Fleet ## Parent #3 (machine-identity) — shares the wire-change conventions and skew gating
• Upstream response in Design: multi-machine pairing (fleet sidebar, per-machine status, rename, per-machine push) KenKaiii/pew2#2 (proceed with the fallback if silent)