Slow backtesting when using Donchian channel

After introducing Donchian channel indicator (DonchianChannel) to my cBot, I've noticed backtesting to be extremely slow. Please find a minimum reproducible example below where I also show elapsed time for each line in OnTick:

using cAlgo.API;
using cAlgo.API.Indicators;

namespace cAlgo.Robots
{
    [Robot(TimeZone = TimeZones.UTC, AccessRights = AccessRights.None)]
    public class BugSlowNestedIndicator : Robot
    {
        [Parameter("Source")]
        public DataSeries Source { get; set; }

        private DonchianChannel channel;

        private MovingAverage T; // moving average of the top line in the Donchian channel
        private MovingAverage M; // moving average of the mid line in the Donchian channel
        private MovingAverage B; // moving average of the bottom line in the Donchian channel
        
        protected override void OnStart()
        {
            channel = Indicators.DonchianChannel(500);

            T = Indicators.SimpleMovingAverage(channel.Top, 1000);
            M = Indicators.SimpleMovingAverage(channel.Middle, 1000);
            B = Indicators.SimpleMovingAverage(channel.Bottom, 1000);
        }

        protected override void OnTick()
        {
            channel.Top.Last(0); // 10'214ms elapsed 
            channel.Middle.Last(0); // 1ms elapsed 
            channel.Bottom.Last(0); // 2ms elapsed

            T.Result.Last(0); // 10'547ms elapsed 
            M.Result.Last(0); // 10'549ms elapsed 
            B.Result.Last(0); // 10'410ms elapsed

            Functions.HasCrossedAbove(Source, T.Result, 0); // 10'395ms elapsed 
            Functions.HasCrossedAbove(Source, M.Result, 0); // 10'335ms elapsed 
            Functions.HasCrossedAbove(Source, B.Result, 0); // 10'120ms elapsed 
        }
    }
}

Questions:

1. How is the Donchian channel implemented? I could not find the source code for this indicator. Using Source.Minimum(Periods) and Source.Maximum(Periods) for bottom and top lines respectively is not the most efficient implementation.

2. Is there anything I can do to speed this up ATM? It looks like the Donchian channel computation is not shared in the moving averages, and thus a recomputation is performed.