Length already isn't a problem for streaming video, the existing video player hack is already limited only by media size.
Bandwidth is the big problem for streaming VBXE video. The video player hack pulls data off the IDE bus at 509KB/sec. The problem is that the only way to get data from the IDE device onto the VBXE would be to copy it with the CPU. That's absolutely the wrong way to use the VBXE as the VBXE can only reach its full potential when it is not dependent on the CPU. Max speed that the 6502 can copy into VBXE is ~250KB/sec; targeting higher resolution/depth with half the data rate won't work without compression, and decompression would have to be done by the blitter. I'm skeptical that the blitter can handle compression good enough to hit quality/size ratio. I've done video compression in this ballpark before, but it was with a 33MHz 68000 and not a simple blitter. In the mean time, the existing video player hack is totally uncompressed, so it has no compression artifacts whatsoever.
In general, streaming tons of raw or nearly raw data isn't a very good use of VBXE, compared to enhanced productivity (multicolor 80-column) or enhanced game graphics. The blitter is fast enough to easily emulate a scrolling tilemap and sprites and it is sorely underused in existing VBXE software. VBXE being an internal add-on makes it relatively inaccessible, though, compared to a cartridge or PBI add-on.