Bug Report

Your environment

• Operating system: Linux or MacOS
• gcc/clang version: -
• mediasoup (Node) version: 3.20.2 (but seems relevant to other versions)
• mediasoup (Rust) version: -
• mediasoup-client version: -

Issue description

For a VP9 SVC consumer created on transport without BWE, when the producer stops sending the top spatial layer (e.g. the encoder drops it under a bitrate cap), the consumer keeps its target pinned to the now-absent layer and never falls back. With K-SVC this means all remaining (lower-layer) packets are dropped and the consumer produces no media at all.

Steps to reproduce

1. Producer: VP9 K-SVC (e.g. L2T3_KEY), 2 spatial layers.
2. Consume it on a transport without BWE (no custom preferredLayers - expect to consume best available layer). E.g. plain transport which connects to ffmpeg to record video.
3. Make the encoder stop the top spatial layer (e.g. cap sender bitrate via RTCRtpEncodingParameters.maxBitrate).
4. Consumer output goes black; no packets are forwarded.

Expected:
consumer falls back to the highest still-active spatial layer;

Actual:
target stays on the stopped layer; in K-SVC every remaining packet is dropped.

Workaround:
For my particular use case I can probably use pipe transport and forward everything verbatim.

Likely root cause

When the top spatial layer of VP9 SVC stops, the sequence stays continuous and no packet loss happen, the producer's RtpStreamRecv score does not change and SvcConsumer::MayChangeLayers() is never triggered.
It normally runs on producer-score changes or on BWE bitrate-allocation events, but on a transport without BWE (e.g. PlainTransport, or any consumer with no transport-wide-cc negotiated) there's no BWE timer to call it either. So RecalculateTargetLayers() never re-runs and the target stays on the missing layer.