You're correct that the talk-sets have to plug into the ends.
However, there's directional indicators that just clamp onto the middle of a fiber. They bend it a little and sample the light that leaks out of the bend, without interrupting payload traffic. The first one I used back in the day was an Exfo but there are tons of 'em now.
As far as I know, these are receive-only, though physics doesn't seem to prohibit launching light into the fiber this way, it would just be an extremely inefficient process.
There isn't enough light leaking out to reconstruct the whole high-bit-rate signal (as far as I know), but there's enough to tell whether the light is flowing one way or the other, or both. And there's enough to tell whether it's modulated with a low frequency signal -- most optical test sets can generate a simple "tone", typically 270 Hz, 330 Hz, 1 kHz, or 2 kHz, and the clamp testers can tell if and which tone is present.
You can't really "get into" an optical fiber mid-run without splicing. Splicing isn't really that hard (I've done it! Fusion splicers are little robotic wonders. Most of the work is in the prep, not the splice itself.)
