brunner 007 - all good advice here, as usual.
My thoughts apply to some matters of historical model development. My relevant experience includes living with the CS2.4, but neither the 2.7 or 3.7. However, I did hear the 2.7 and 3.7 in the Thiel listening room the day the finalized 2.7 arrived from the Canadian outsource engineering firm, Warkwyn. First-off, I consider the 2.7 to be a 'real' Thiel speaker. Although developed after Jim's death, home team Thiel faithfully applied their work to Jim's vision, methods and standards. Multiple circuit changes were authored by Warkwyn, but coaching and approvals were from Thiel. The results prove their good work.
In direct comparison, the 3.7 is the more articulate speaker with deeper bass and greater precision and dynamic range. But the 2.7 holds its own. Your room, amplitude and budget should guide you more so than differences between the models. Note that the 2.7 uses the bass driver and alignment of the 2.4 and the coax of the 3.7. XO changes are required for the cross-point changes between the models.
That said, I'll add that the model 2 has a sonic signature with a somewhat elevated mid-bass. That came about in the CS2.2 being our first passive radiator, coupled to the room better than anticipated. The bigger bass was pleasant, plus many listeners liked it better than Thiel's more accurate target. We kept that signature for the model 2. (I've never heard the 2.3, and the CS2 is beyond accurate memory.) But I have used the 2.2 throughout my career as monitors, and the 2.4 resembles its balance. Perhaps the 2.7 might too.
Similarly, the 7.2 (and I believe the 7) have an accentuated deep bass. Impressive, but somewhat less than technically accurate. I suggest you consult the Stereophile reviews for your products of interest. Also, both cables and especially amps can produce such shifts in tonal balance. The speaker is part of an interacting puzzle.
Beyond all that, if the 2.7 lands on your radar, we have a simple upgrade that cures its largest shortcoming. Warkwyn specified a single 400uF electrolytic series feed cap. Jim never exceeded 100uF to keep parasitics down. He would have used 4x100uF caps. I use 8x50uF in a cylindrical array around a new 1uF cap made of 10 cascading sections. A part-swap would elevate performance substantially.
Note that the CS3.7 is functionally a 3ohm load, as is the CS5 for purposes of driving. Only a few amps in the world can drive them to their potential.