Hey @Chris_Gosling,
Thanks for writing in and for sharing your report! From a fresh Roon Server diagnostic report, we can see that the LSX II in particular drops to standby repeatedly mid-track, seconds after playback begins, regardless of format (FLAC, MQA, DSD). This is happening on both speakers but is far more frequent on the LSX II. This is the primary cause of the crashes and inability to activate the endpoint, Roon starts playing, the speaker reports back status: Lost, reason: standby, and the endpoint goes dead.
Based on this, and a few other reports from Kef releated issues, I believe this is a firmware bug in the new KEF firmware. The speakers are entering standby erroneously while Roon is actively streaming. The KEF Connect app bypasses Roon’s RAAT protocol entirely and uses the speaker’s own internal playback path, which is why it works fine there.
We’re also seeing that both speakers are internally downconverting everything.
Both speakers still advertise full DSD support (DSD64/DSD128/DSD256) to Roon, so Roon tries to send DSD natively. However the signal path logged shows what actually happens:
{"type":"dsd_to_pcm","from_sample_rate":11289600,"to_sample_rate":192000,"quality":"none"}
Every single DSD setup request whether DSD64, DSD128, or DSD256 results in the speaker internally converting DSD to PCM (192kHz for LS50 II, 96kHz for LSX II). The quality: “none” flag means this is lossy/degraded conversion happening inside the speaker firmware, not the hi-res native DSD path you’d want.
This is new behaviour also seems to be introduced by the recent firmware updates. Previously, these speakers passed DSD natively; now they’re forcibly converting it on their end.
As a next step, if you can, I’d try to roll back KEF firmware. The LS50 II’s 4.1.190→4.1.191 update and the LSX II’s 3.0.135 (compiled April 20, 2026) are the clear culprits.
If KEF offers a previous firmware version, downgrade. Check KEF’s community forum, you are almost certainly not alone. Our partners team has reached out to KEF directly around the recent firmware-related issues.
You have our apologies here, Chris! 