I had two micros when I was growing up: the SAM Coupé and the Acorn Electron.
The SAM Coupé has a 512x192 mode much like you discuss. The SAM's display is physically wider than a Spectrum's though, pixels being approximately 1.25 times as wide as tall.
The Acorn Electron inherits full 80-column 640x256 video from its progenitor, the BBC Micro.
In both cases my recollection, across the TVs we had in the house during the early '90s, is that the high-resolution text was legible but uncomfortable. I spent a lot of time programming on the SAM in particular (because it's so much nicer when there's a floppy drive) but don't recall spending all that much time in its high-resolution mode.
To go beyond the qualitative, it just so happens that I'm an emulator author and that one of those includes an emulation of the Acorn Electron and is capable of displaying its output via a composite signal. This isn't one of those hacky "PAL filters" that tries to post-process a frame according to what I and an echo chamber feel like a TV did, it really generates a real composite signal, then really decodes it. So colour artefacts, frequency limits, occur naturally and accurately. Apologies for the sales pitch, but I hope it explains why I think the screenshot below is relevant to the conversation — that's what happens when I simulate the Electron's 640-pixel video mode via a composite signal.
Honestly, it looks worse than it is when pictured statically, as the PAL colour clock is out of phase with the pixel content, so it sort of swims frame by frame: the colour fringing moves around and your brain is able to get a little more of the form than it can from just a static shot.
NTSC is even more limiting than PAL: due to a different trade-off between available spectrum and quality of signal, the luminance part of the signal can contain only around 80% as much detail.
The Apple II, which acquired an 80-column mode not too long after launch, gained a quick fix for that: in text mode it declines to provide a colour burst. The colour burst is the part of a composite signal that tells a TV both the the current colour phase and indicates that this is a colour-encoded signal. High-end sets born into a world where they might receive a colour or black and white signal can use knowledge that what they're showing is black and white to apply less filtering to the signal. But all sets should decline to decode colour, so you should at least be spared the colour fringing.
Also, cheap black and white sets will usually give a much better picture: they make no effort whatsoever to remove the colour subcarrier. You get a full-frequency image. If the source image is black and white then that can be exactly as good an image as a real monitor. If it has colours then you'll see some chroma dots around, but both PAL and NTSC subcarriers were selected to reduce their visibility: that's why they're out of phase with the scanning frequency.