Current matroska/webm-mux create BlockGroup with timestamp and duration for each vorbis block. IMHO it is a useless overhead.

Background:
Usually Blockgroup with duration used together with DefaultDuration to reduce overhead. In case with vorbis we can't use DefaultDuration.
Vorbisencoder use variable block size. It variate from 64 to 8192 samples per block. What minimal and maximal blocksizes will be used, depends on initial settings (like sample rate and quality/bitrate), and some other decisions made after start for specific sample groups.
After codec start we can ask minimal and maximal blocksize, and probably use some middle one to calculate block duration, but this behavior may variate from one codecversion to other.

So we can't use DefaultDuration.

About the use of BlockDuration. Since we know samplerate and probably can read current blocksize of vorbis in blockheader, there is no sense to use blockduration.

In case the blockduration is less as difference between current and next timestamp, then we play silence in the rest of the time. If it is bigger, then some thing is wrong and muxer is brocken. But we can't fix it if we write blockduration.

Are there any other use for blockduration for vorbis? If not, it should be removed.