Only with filters that :
* Dont need several frames (e.g. deinterlacer)
That's quite a bad change compared to the 2.5 branch - I process a lot of movies recorded on TV, and quite often those have 'sped up' end credits, having only those interlaced - now I have to deinterlace the whole video instead of only the last 1-2%.
Although it doesn't 'hurt' the non-interlaced part, it's unneccessary filter processing being applied to the huge main part of the video.
What's the technical difference between deinterlacing of the whole video from frame 0 to End (I don't have a frame -1 and End+1 there...) and processing only frame A (>0) to End?