#299 OPERATOR PICK: (A). Build it after your #300 diff is in, same product branch, its own commit:
- PRESENCE_PROBE_CEILING parameter 10000 -> 2500 ms (crates/spt/src/cli.rs ~6968). The shared PROBE_TIMEOUT const stays untouched; the ceiling stays a separate parameter that happens to carry the same value. Doc comment: the justification is this measurement — perri's 10105/10120 ms were one unreachable peer's dial reaching the ceiling; a probe past 2.5 s reads UNKNOWN, never OFFLINE, exactly as before past 10 s.
- REQ-UNLISTED-PRESENCE-PROBE title: "HARD 10-SECOND" -> 2.5-second, with the #299 measurement named; CONTEXT.md presence-probe entry sentence likewise. Corrected by replacement.
- Re-pin the existing 'wedged probe costs ONE ceiling' cell to the parameter, not a literal. Any other test asserting the 10 s literal: grep and re-pin (behavior -> grep tests).
- Closing comment on #299 carries: the attribution, the 457 ms pre-subnet block, the 30 s dial vs ceiling note, and this follow-up the operator stated: presence should ideally be PUBLISHED by heartbeats/peer pump so that list reads recent cached presence even after a stall — filed as a separate request, not built here.
Targeted cells + traceable check; no full suite.