It seems that most of the modern radio astronomy instruments and observation that make the news are interferometers or phased array systems of one kind of another. Is there any application left for which a single huge dish is better (or cheaper) than the same collecting surface in the form of some kind of array of smaller dishes, whether relatively compact, like ALMA, or widely distributed like the Event Horizon telescope or the square kilometer array? In other words is there any point at all in considering replacing Arecibo by anything the same size?
Single-dish telescopes have advantages over interferometers in a few areas; existing answers have touched on some of them. Collecting area is extremely important, as Rob Jeffries mentioned, and you need extremely large arrays to compensate for this. Granted, such arrays are certainly possible (ignoring the fairly sizable cost cost), as demonstrated by the coming Square Kilometer Array in 2027, which will have a collecting area of ~1 km. On the other hand, the SKA is in many ways the exception, not the rule, collecting area included.
Another advantage of single-dish setups arises when investigating large-scale structures. An interferometer's upper limit to its spatial frequency depends on the shortest baseline between any two dishes. Even in its most compact configuration, the Very Large Array's minimum baseline is 35 meters. If you want to quickly map large areas of the sky - important for large, extended sources - you want a single-dish telescope. In cases when you require both large-scale and small-scale spatial sensitivity, a combination of an interferometer and a single-dish may be warranted.
Let's say you want to swap out receivers, or upgrade your instrument for a specialized purpose. In that scenario, it's much more convenient to just have one dish to deal with. It would be much, much simpler to install a new receiver in one dish instead of dozens. Similarly, an interferometer won't be able to easily shift between configurations, which gives an advantage when it comes to scheduling - particularly in an era when radio telescopes are often massively oversubscribed.
As a final note, from a spatial perspective, observations of point sources - pulsars, FRBs, radio transients, etc. - can be done just as easily with single-dish telescopes - which, again, usually have larger collecting areas.
Arecibo wasn't just a radio telescope, it was a radar telescope, bouncing megawatt-level radio signals off various bodies in the Solar System. A single-dish transmitter is far superior to a phased-array or other composite system, because the beam pattern is a simple Airy disk rather than a complicated pattern formed by tens or hundreds of such disks.
Having a large dish gives you a large collecting area and hence better sensitivity. Building a multitude of receivers with the same collecting area, each having its own feed and electronics, is more expensive, not less. Otherwise that is what people would have done in the past. Arrays are built because you can synthesize a larger diameter of aperture. In addition, the noise properties of 100 receivers, each with 1/100 of the collecting area of a big dish do not give the same sensitivity, because you will essentially be getting additional noise for each receiver that is added.
I've read of plans to build a large single disk radio telescope somewhere on the Moon, to give access to wavelengths blocked by Earth's atmosphere. It would be best to choose a place that gets no radio signals from the Sun or from the Earth, which would probably be in a crater near either the Moon's north pole or it's south pole. Such a location could not see the entire sky, so a second one may be needed on the opposite side of the Moon.