Thermocouple effects will overwhelm the low drift of either solution, or the LT1028 or similar precision amplifier used alone, so the noise difference over your bandwidth of interest is all that matters. You can most likely get away using just a modern low noise chopper stabilized operational amplifier. The LT1028 solution would be required for lowest noise when making measurements with a bandwidth of more than a few Hz.
With the 4 wire connection, there is a common mode voltage to be removed, but the highest common mode rejection will not be required. I would pay more attention to how you plan to do this.
Long ago I extended your example of the LTC1052 correcting the offset and low frequency noise of the LT1028 to a fully differential configuration and got low noise performance which was difficult to imagine. It worked so well that I could measure low values of resistance by Johnson noise alone to within a couple ohms. (1)
(1) The fun but unsettling part was calibrating the breakpoint frequency between the chopper stabilized amplifier and the operational amplifier for lowest noise. I found a marvelous solution for this, which this post is too narrow to contain. (2)
(2) I was just kidding. I used a high resolution digital voltmeter in sampling mode and a calculator to measure and calculate the 0.1 to 10 Hz noise, while adjusting the time constant for minimum noise. Once the various gain factors were accounted for, the frequency breakpoint was exactly where the noise curves of the LT1028 and chopper stabilized amplifier crossed, which was reassuring.
The main intention of combining the LTC1052 with the LTC1028 is to overcome the bandwidth limitations of AZ amps.
Someone might have done that, but I largely disagree because chopper stabilized amplifier are usually pretty fast anyway, and the aliasing problem occurs when the input signal content is close to the chopping frequency.
The purpose is to limit the input offset voltage drift and rising flicker noise of the LT1028 at low frequencies.