Fastest and/or most reliable way to connect to BLE central/server

Hi,

sorry for opening a new topic again already, but I hope that our notes might help somebody else later on.

Right now I am connecting like this to my bike sensor:

NRF.requestDevice({ timeout:5000, filters: [{ services: "0x1816"  /* namePrefix: 'Wahoo' */ }] }).then(function(device) {
  console.log(device);
  return device.gatt.connect();
  }).then(function(gatt_object) {
    gatt = gatt_object;
    s = "Looking for service...";
    g.clear(); g.drawString(s ,0 ,0); g.flip();
    return gatt.getPrimaryService("1816");
    }).then(function(service) {
      s = "Looking for char...";
      g.clear(); g.drawString(s ,0 ,0); g.flip();
      return service.getCharacteristic("0x2A5B");
    }).then(function(characteristic) {
      characteristic.on('characteristicvaluech­anged', OnNotify);
      return characteristic.startNotifications();
      }).then(function() {
        connected = true;
        t = getTime();
        setDisplayMode("livedata");
        gatt.device.on('gattserverdisconnected',­ function(reason) {console.log("gattserverdisconnected, Grund: "+ reason+" Zeit: " + (getTime() -t) + " sec"); disconnect(reason);} ); 
}

Initially, I even had a timeout of 20 seconds in requestDevice, which led to pretty unreliable connection intitation.
With nrfConnect I see that the sensor starts advertising really fast with an interval of 330ms initially.
I wonder if it is possible to cut the timeout short (right now it seems that the code scans for the full timeout before handing back scan results) and return a matching device immediately?

Or is it preferable to cache the sensor's address and do a direct connect to that later on? That would have the (very small) drawback of needing additional logic to do a full scan if the previous/cached device is not successfully connected, to enable connecting to new/other sensors.

Thanks,

Joost