Ears is mainly a contact mic module. Its main point is its built-in contact mic for creating bonks, scrapes, etc. As an external input module, it's missing lower impedence settings (1 Mohm only), and requires jumper changes to change its preset attack/release times and input sensitivity level. Contrast that with something such as Doepfer's venerable A-119: variable levels and thresholds on the front panel with no presets, and two different inputs for either line-level signals or low-level instruments or microphones on appropriate jacks. As someone who prefers my contact mics on the other end of a cable, this seems to be the convenient option.
As for ES-3 vs ES-8...yes, you want USB and you definitely want the four return channels. USB is much easier to deal with in many settings, whereas a lightpipe-only interface like the ES-3 largely anchors your modular to a fixed position that has an ADAT lightpipe I/O somewhere. And those four returns aren't just for recording your system's output; they can also serve functions such as allowing the modular to work as your session clock, sending triggering signals from the modular back into the DAW to allow the synth to determine actions in software, sending modulation from the modular back into the software for use there, etc.
Last, cables. Get them. Get LOTS of them. If you think you have enough cables, you probably don't have enough cables. And use inline mults; 126 hp x 2 is still small enough that I'd recommend not using passive mult modules in the case in order to squeeze a little more function in. And buffered mults really serve their purpose when you've got to spread a CV/gate/trig out across 5, 6 etc modules; just splitting a CV between 2-3 VCOs, for example...that doesn't need a buffered mult with the vast majority of VCO modules these days.