Function doubts_now 

pub fn doubts_now(
    due: i64,
    now: i64,
    record_lost: Option<i64>,
    peers: &[Peer],
) -> Vec<Doubt>

Why a dose due at due (Unix seconds), not answered here, may have been answered on another device, as of now; none when it is known not taken (docs/health.md, “Knowing”). This device’s record must read, and each other device that counts (counts) be one of:

• closed cleanly (working down, closed after started, both on its own clock), everything it wrote up to its last export read here: closed after the dose, or before it and not started since. Its last export holds every answer it captured, and it captures none until it starts again;
• in use, having exported after the dose (its clock, allowing SKEW and how far ahead its clock was seen) and lately (seen changing here within FRESH, and dated within it), everything read: an answer captured since goes out at once and comes within the sync’s time;
• an older Sioul, which writes no registry: as its claims say, heard in full after the dose and within FRESH, or its last word, read in full, that it closed, its news quick (NEWS_IN). Never closed for want of a word.

Otherwise the doubt names it. One rule for reminders, near the dose’s time, and for what stays in view hours after it (the Porch, the Health page, a phone’s home screen). A dose taken twice can harm: what is not known is said, never guessed.