Not sure what you mean by layers and key ranges...

A SCENE is snapshot memory, it can recall: Arpeggios, Motion Sequences, Super Knob setting, Mix setup incl. Rev Send, Variation Send, Dry Level, Pan position, Volume, Filter Cutoff, Resonance, FEG Depth, Mute status, ADSR, Arpeggio and Motion Sequence Play FX.

If you need to store different sounds layered with different Key Ranges for each that would not necessarily be a SCENE but a job for the Live Set, or even the Super Knob (since both allow you to seamlessly change between different configurations)

Do scenes have seamless sound switching?

Seamless Sound Switching (SSS) is a function of Performance mode and Live Set.

So there is not a definitive yes or no answer. If you need switch between several sounds the "snapshot" memory is not necessarily what you would use for this... Mute/unMute is not seamless as in having once sound sustain into the other, neither is Volume going to zero. Scenes are for mix status changes.

Better options would be Assignable Function buttons controlling XA CONTROL (Expanded Articulation), the Super Knob and its ability to morph between sounds, or simply place the sounds you need to move seamlessly between in a LIVE SET.

Although this is a synth so I guess you could have Scene 1 with one sound, Scene 2 with both sounds and Scene 3 with other sound. But there are simply other better ways to do SSS.

So changing scenes while playing will often generate sudden/abrupt changes? (like volume, cutoff, pan, resonance)

If you're holding a chord and press Mute, the sound could not stop quicker. Mute is not what you want, you want to Morph, or have both sounds overlap...

Don't over think it, let's use visuals as an analogy. A Scene is considered "snapshot" memory because it is like a photograph. The term literally comes from the 1970s recording studios. Back in the day before studios had automation, every studio had a Polaroid SX70 Land Camera, you know the one that spit out the picture with the chemicals on the paper, and you air dried the photo; in about 60 seconds you had a picture.... Because if you needed a backup of your console settings, your patchbay, you took several photos and kept them in the box with the two inch tape. Grease pencil marks never survived through the next session.

Snapshot memory. Switching pictures is never smooth... In visuals, a crossfade is smooth, morphing between controller positions (on a continuous controller) would be smooth. The Expanded Articulation function is designed for seamless switching between Elements you are performing. Scenes do not always cause an abrupt cutoff, however, because a Scene change can include a message to, for example, switch Arpeggios and/or switch Motion Sequences at the top of the next measure, so it transitions the instruments and rhythmic Parts in a musically appropriate manner.

If I'm reading your question about 'seamless' correctly I envision you are talking about holding a chord with the first sound, changing to a second sound and having both overlap. For that type of seamless transition... I would not necessarily use the Scenes for that... There are so many other ways, ways specifically designed for that very purpose. Why use a Scene, when that is not it's designed purpose?

All of the Performances come SSS already so nothing extraordinary needs to be done to hold one sound, either with your hands directly or with the Sustain pedal, when you call up the next Performance you can immediately play it. The original sound will hold until you let go... If you are holding the keys down, when you release them, if you used a sustain pedal to hold the first sound, seamlessly switching to the second gives you full access to all the keys.