Bug: Watcher sends duplicate query at startup

[MoonBoi9001]

Bug: Watcher sends duplicate query at startup

Summary

The new_watcher function executes its query function twice at startup due to Tokio's time::interval first tick completing immediately. The second query is redundant, and its result overwrites the first - so a transient failure on the second query can replace valid data with empty/incorrect results.

Observed Behavior

Debug logging confirms two identical queries at startup:

23:23:40.272Z DEBUG: get_allocations called, closed_at_threshold_secs: 1766877820
23:23:41.233Z allocations: N
23:23:41.234Z DEBUG: get_allocations called, closed_at_threshold_secs: 1766877821
23:23:48.767Z allocations: N

Both queries use the same parameters (1 second threshold difference is negligible on a 30-day buffer). However, in earlier testing the second query intermittently returned 0 allocations instead of the correct value, overwriting the valid first result and causing receipt validation failures. There is a PR to fix this here #912

Root Cause

// crates/watcher/src/lib.rs
let initial_value = function().await?;  // Query 1
let (tx, rx) = watch::channel(initial_value);

tokio::spawn(async move {
    let mut time_interval = time::interval(interval);
    loop {
        time_interval.tick().await;     // Returns IMMEDIATELY on first tick
        let result = function().await;  // Query 2 (redundant)
        tx.send(result)...              // Overwrites initial_value
    }
});

[MoonBoi9001]

The Gateway intermittently returns 0 allocations for valid queries.

In one restart, the first query returned 0 allocations while the second returned the correct count. The reverse also happens though.

Any fix should probably be paired with validation logic for unexpectedly empty responses.