[Feature Request]: Configurable Node Pruning for Stale/Ephemeral Nodes

[Syridian]

Summary

Add an optional feature to automatically prune stale nodes from the NodeDB to improve routing efficiency and reduce clutter in busy meshes. This is especially useful for fixed routers that should ignore transient or “drive-by” nodes.

Why This Is Needed

Routing stability: Ephemeral nodes cause frequent route recalculations and wasted airtime.
Flash conservation: Prevent NodeDB from growing indefinitely.
Mesh hygiene: Keeps routing tables clean and predictable.
Router specialization: Allows backbone nodes to focus on reliable participants.

Proposed Behavior

Nodes are pruned if:They have been heard ≤ prune_min_hear_count times (default: 1)
They have not been heard for prune_hours (default: 50 hours ≈ 2 days)
Failsafe conditions:Skip pruning if NodeDB has fewer than prune_min_nodes (default: 5)
Skip pruning during first prune_grace_hours after boot (default: 6)
Router protection: Do not prune nodes with ROUTER role by default.
Schedule: Run pruning once per day during housekeeping.

Configuration Interface

Expose new fields in DeviceConfig (protobuf):

bool prune_stale_nodes = <id>;      // Enable/disable pruning
uint32 prune_hours = <id>;          // Time threshold in hours
uint32 prune_min_hear_count = <id>; // Minimum heard count
bool prune_skip_routers = <id>;     // Protect routers
uint32 prune_min_nodes = <id>;      // Failsafe minimum node count
uint32 prune_grace_hours = <id>;    // Grace period after boot

CLI examples:

meshtastic --set prune_stale_nodes true
meshtastic --set prune_hours 50
meshtastic --set prune_min_hear_count 1

##Short Code Snippet
Here’s an example of where pruning logic hooks into housekeeping:

void MeshService::housekeeping() {
    static uint32_t lastPrune = 0;
    uint32_t now = getSeconds();

    if (now - lastPrune > 86400UL) { // Run once per day
        lastPrune = now;
        nodeDB.pruneStaleNodes(deviceConfig);
    }
    // ... existing housekeeping code ...
}

Benefits

Keeps NodeDB clean without breaking routing tables.
Fully configurable for different mesh types (public vs sensor-heavy).
Safe defaults and failsafe mechanisms prevent accidental over-pruning.

Source

This feature request was a collaboration using ChatGPT, conversation PDF available on request.

[Syridian]

No feedback? No comments? Bad idea?