First of all, Fiddler is a spectacularly useful tool and I thank you very much for providing it.
Yes, most textual web content is encoded in UTF-8, but I would venture to suggest there is as least as much non-textual web content (from .jpegs to .zip files). And yes, trying to make sense of arbitrary binary content as text is doomed. It would be nice if feeding the Raw tab such data did not cause the UI to hang.
However, when I'm faced with unknown content that might be text, the raw tab is where I want to start (even .zip files have snippets of decipherable ASCII). A quick glance at the bottom of the pane distinguishes text from gunk (unlike the HexView). The headers are often important, so "Raw" should be perfect. If it has western text, then it is easy to decipher. If binary, one can still read the headers (unless some random NUL has stopped things up) and the rest is gunk, sometimes with some readable text blocks here and there. One can go over to the HexView to see the particulars of the binary content and see the very same Latin1 characters in HexView's third column but with the original line endings lost and irrelevant data in the other columns taking up most of the pane.
Random data (mis-)interpreted as UTF-8 or UTF-16 can get at every character the system can render, one can wind up with a text box containing both right-to-left and left-to-right text, every weird character one can think of (from symbols to runes), one winds up giving the text rendering engine and CPU quite the workout. Latin1 has less than 256 possible symbols and everything renders left-to-right. That is Western-centric, but protocol-related text usually is (e.g., HTTP headers and the "Content-Disposition: form-data;" stuff from above); such ASCIIish stuff encoded as UTF-8 is identical to the Latin1 encoding. If there is non-Western text, then one usually needs to get the encoding right to get anything other than mojibake. More fundamentally, there is already a TextView tab that is there for showing stuff known to be text (a box to force the encoding to an arbitrary entry in Encoding.GetEncodings() and a quick way to switch to a fixed with font would be great there).
Try playing http://devimages.apple.com/iphone/samples/bipbop/bipbopall.m3u8
with VLC while monitoring the traffic in Fiddler. It is not a live stream, so there are no playlist downloads interleaved between the media segments. With the auto truncation disable, the Raw tab works great for looking at the playlists together with their headers (the cache-related headers are often important and while the content of the media segments do not matter in this context, the headers and the order of the segment GETs with respect to the playlist GETs is important). Then click to a media segment and back to a playlist.
TL;DR There is a tab in Fiddler called "Raw" where if one accidentally clicks on the wrong session, the UI will freeze for a very long time.
Thanks for your time.