Let me say that the QuickTime/MPEG4 container sucks. It has scalability problems because the structural and informational data atom 'moov' and the codec data atom 'mdat' can not be interleaved. A compressor must write to separate moov and mdat files until the end of the audio/video is reached, append the mdat atom to the moov atom, then re-index the moov atom to reflect the new data offsets in the mdat atom. Alternatively, the compressor can pre-allocate some space for a moov atom and hope it doesn't run out before compression finishes. A decompressor must buffer all moov atom before it can play anything from the mdat atom. In other words, it's not actually streaming. It's a total pain in the ass. This is why video cameras producing MPEG4 files are limited to a few minutes of operation at a time. This is why phones have trouble playing long QuickTime movies even if they have hardware acceleration.
The real kick in the nuts is QuickTime X. It was Apple's chance to drop all of the legacy baggage but the first step was using the ancient QuickTime container. There's nowhere for it to go now. It can't properly support streaming files like MPEG2 or AVCHD.