"Why would you even want that?" questions aside, from a programming perspective there's a very simple reason: the OpenType spec only affords an addressable glyph index space of one USHORT, so one font can only support 16 bits worth of glyphs identifiers, or 65,536 glyphs max. (And note the terminology: a "glyph" is not the same as a "character" or "letter")
The current version of Unicode, v8 as of this answer, contains 120,737 assigned code points, or almost twice as many as fit in a modern font (2021 edit: v13 upped this number to 143,859). In fact, Unicode hasn't been able to fit in a modern OpenType font since 2001, with the release of Unicode 3.1, which upped the number of code points from 49,259 to 94,205.
"So what about font collections?" I hear you ask. Why not use multiple fonts and support all unicode that way? Well now, you've just described Adobe's Sans Pro, and Google's Noto (which are the same font).
As for the "how hard can it be": a uniform style for all glyphs in Unicode, across 129 established written scripts on this planet, each with their own typesetting rules? Incredibly hard. You may think fonts are just files with pictures for letters, and someone types a letter, that picture shows up: that is not how fonts work, and isn't how fonts have worked since the late 1980's.
Modern fonts are the typographic equivalent of a game ROM: sure, it's not much use without the hardware or software to run that ROM on, but all the things that actually matter are in the ROM. Similarly, modern fonts contain all the information for typesetting. Not just pictures, they contain the metadata, the metrics, the positioning and substitutions rules for arbitrary sequences, with separate rule sets for each written script that OpenType supports, mandatory and optional ligatures, language-specific character replacements for letters at the start/middle/final position in a word, or in isolation, character repositioning relative to arbitarily complex sequences of other characters either before or after it, arbitrarily complex sequence replacements with other arbitrarily complex sequences, possible bitmap fallbacks for small-point rendering, hinting instructions on how to properly rasterize vector graphics that are inherently not aligned to any particular pixel grid, and more. A modern font is a ridiculously complex application, that a font engine consults to figure out how to typeset sequences of code points.
Making a (set of) Unicode-encompassing font(s) that looks good for all contexts is a vast team effort.
So: "Why isn't there a font that contains all Unicode glyphs?", because that's been technically impossible since 2001. We can, and do, make font families that cover all of Unicode, but with 129 different scripts all with their own typesetting rules, it's a lot of work, and almost (almost) not worth the effort compared to only covering a subset of all languages.
And as for this:
Such a font would be a great help to humanity and I can't see a good technical reason why it doesn't exist or at least an open-source effort to create it, so I presume an invisible-to-me reason why it can't be done.
Just because you didn't know about them, doesn't mean they don't exist, with millions of people who are familiar with them. They exist =)
They're even open source, go out and thank the people who made them!
