This matches how Lumen’s final gather works: with more allowed bounces, the per-sample ray budget gets split across more bounce segments, so each individual bounce (including the first) receives fewer/shorter traced samples and the result converges darker before it gets better.
Worth checking alongside maxBounces:
r.LumenScene.Radiosity / downgrade of gather quality — raise r.Lumen.FinalGatherQuality when testing multi-bounce so the comparison is fair.
Emissive contribution: multi-bounce lighting from small bright emissives needs a higher r.LumenScene.EmissiveInjection step; at low sample counts the extra bounce can be lost in noise and the temporal filter averages it down.
Compare with a fixed seed (r.Lumen.TraceMeshSDFs 1, disable temporal jitter via r.Lumen.HardwareRayTracing.TestDataPath if you’re on HWRT) so you’re not comparing different noise patterns.
If 1-bounce genuinely looks more accurate rather than just brighter, that’s likely a bug worth reporting with a minimal scene — but first rule out sample-budget dilution by comparing at matched final-gather quality.
