This is going to sound snooty, but imo the more you know about music theory, the more underwhelming the scale and chord features of various sequencers are going to be.
Let’s say a sequencer can produce every possible chord in every possible inversion. By the time you deal with all the menu diving and/or knob turning required to choose one of a few hundred chords, it’d be easier to just enter the chord manually, one member at a time. That is, if you know how to spell a chord.
To perform a scale manually, you must first know the scale. On the other hand, to perform the same scale from a machine preset, you have to remember where that scale exists on the machine.
If we spent the same amount of time learning music theory and practicing scales as we do learning to navigate our machines, some of the gas we have for the so-called “smart” features on the latest, greatest devices…would evaporate.
The fact that unadorned scales and modes have 7 scale-degrees, creates a problem when these scales and modes are used to fill even numbers of beats. Octatonic scales will make more effective fills, repeating at the octave over an even number of beats. Similarly triads, when arpeggiated, do a poor job of filling even numbers of beats. A fourth member of the chord, or a non-chord tone, must be added in order for the arpeggio to replicate itself in a more predictable fashion.
Maybe the kind of sophistication I’m describing already exists in some sequencers, and I’m not aware of it. My frame of reference for wanting features from a sequencer is based on knowing (or thinking I know) exactly what I want to do with it. Too many features of too much gear, by contrast, seem to be based on chance and randomness.