Thanks for asking. Playback compression is very different from Replay Gain.
Replay Gain, stated simply, tries to make all tracks play at the same volume, by raising/lowering the entire track's volume relative to other tracks, while avoiding clipping any track's peaks. It is one adjustment applied to an entire track.
Dynamic range compression does more, differently, for a different purpose. It adjusts audio level, millisecond-by-millisecond, as the track plays. There are many ways to tweak this, but the general effect is that low parts of the song are raised and loud parts (including peaks) are lowered. In broadcasting, the lowering of peaks then allows the entire track's volume to be raised, making the radio station's usable signal go a bit farther. AND, compressing of dynamic range makes the entire song sound "louder" because our ears react to average sound levels, making the radio station jump out on the dial.
Compression of ALL songs creates a more consistent, controlled sound. Leveling out the volume range and maintaining a consistent average is a huge help in noisy environments such as moving vehicles, outdoors, bars, clubs, and even hotel lobbies and elevators. PA systems and commercial amplification systems usually have built-in compression.
At the song level, dynamic compression creates a hotter sound. People who grew to love a song on the radio often wonder why it sounds wimpy on the CD/album, and the reason is the added compression of the radio station.
Of course, compression is used on every recording, whether noticeably or not, to keep the audio above the noise level and below the clipping level. Because this is applied by the recording/mastering engineers, it's not consistent among recordings. Maybe across a single album, but often not even across a single recording session, and definitely not across multiple albums or miscellaneous tracks.
Dynamic playback compression can do quite a bit to clean up the wide variations in recording/mixing/mastering, creating a much more consistent sound as multiple songs are played. This is a big reason radio engineers see compression as an art; getting it just right is point of pride.
Of course, the compression parameters will vary depending on a station's musical format. I used very light compression at a classical music FM station, more on the same station's AM side, but a lot of compression on rock stations, AM or FM. Without the compression, is just sounded "wrong".
Taking this perhaps to the max, radio stations now use devices that have several compression channels in a multi-band process. The incoming audio is split into separate frequency ranges, typically five (loosely, bass, upper-bass, low mid-range, mid-range/vocal, upper mid-range, highs). EACH range is compressed by itself. Then the result is recombined into the outgoing full audio. The result is amazing, essentially a dynamic re-mixing of the original track. Bass-weak songs (like the Beatles on Capital) get "bottom". Screechy recordings get toned down. Tinny or muddy songs become clear and vibrant. I've heard it create reasonable consistency across recordings spanning 60 years.
Of course, compression can be misused. The common complaint that TV commercials are "too loud" is simply because they are compressed more than the show's audio. (When I produce TV spots we always compress the audio because that's what the advertiser wants.) Another complaint is that theatrical movies on TV (on HBO, for instance) are hard to hear in the family room because films intended for a big sound system in an otherwise quiet theater use much less compression than shows produced for TV. Smarter TV broadcasters tweak up the sound of movies, but some go too far, given that a mix of dialog, music and effects on TV should not get the same treatment as rock 'n' roll on the radio. Some cable channels I watch get it wrong a lot: AMC, Sci-Fi, Encore, often drive me crazy. (Not to be confused with the fact that all cable/satellite systems control the audio level of every channel individually, and sometimes do it incompetently, the reason as you channel surf some are too low relative to others -- call and complain.)
The same happens inversely. Lack of audio compression is why web sites that try to play music or present audio, people who record podcasts and vidcasts and "home" movie/video productions, often sound lame. They don't know about or effectively can't apply audio compression (and its partner EQ).
Back to MC: When playing music locally, not via a radio station's "audio chain", a variety of tracks in a stream can have jarring volume variations. Replay Gain is the cure for this, preventing wild swings in volume. But some songs will still sound louder than others, even if Replay Gain has done everything it can to make them the same. That's because of variations in the mix of different songs, and a big part of the mix is the compression.
(The critical factors in mixing a recording session's multiple are track EQ, track relative levels, and track compression -- critical because they interact and a problem with one can't be cured by diddling the others. And at the head of the "get it right" list are the original recording session decisions of microphone characteristics vs. live sources, and mic placement vs. sources vs. room acoustics.)
On the listener end, where we have a collection of finished tracks and can't remix from studio masters, dynamic playback compression is a terrific tool to create a more uniform "sound". But -- It's definitely a matter of taste. Some people (me) love the sound of a well-engineered radio station and lament the thinness of the same music on CD. Others don't. That's why, if it ever becomes available, it should be optional.
Re albums vs. tracks, that's personal preference. I love variety and rarely want to listen to an album as such. Random is my favorite mode. My musical collection spans almost every genre and year that audio has been recorded, probably because I'm a musician and get called upon to play all kinds of stuff, and probably because I have a short attention span.