I'm not a signalling engineer but I asked a colleague about this a few years back when I noticed the same. IIRC he stated that it's got two purposes:
1) (Depending on how close signals are together) If the train following behind sees the signal at green and empty track between them and it, the driver may focus on that and SPAD their own signal which, of course, will still be at red because the green is only a distant.
2) If a train breaks down or has to stop for an emergency or something, it will often deliberately stop at a main controlling signal to allow for easier contacting of the signaller and location identification. If the failure or emergency requires another train or loco to come and assist, and this comes from behind the failed train, the assisting driver will receive a green aspect followed by the rear end of a train, and thus the chance of a collision is increased. Whilst, being a distant, it's unable to show a red, a yellow is at least more restrictive, and thus if the assisting driver does loose situational awareness, they will at least be going slower, expecting a red, so when they find a train's rear end they are much less likely to crash into it.
Thus, a separate track circuit is usually provided specifically to trigger this reversion to the most restrictive aspect the signal is able to display. It can be omitted by risk assessment I believe, but it's general best practice to include it.