IBM have recently announced their 127 qubit Eagle processor. Other approaches, such as Rydberg arrays, have now 256 qubits, as for example in QuEra's QPU QPU.
While these are without a doubt outstanding techical acheivements, I am wondering what is their intended use case. The reason I'm asking is that as far as I know, a rough estimate that the number of qubits that can be entangled without error correction is about $1/\sqrt{\epsilon_{2Q}}$, where $\epsilon_{2Q}$ is 2 qubit gate error. Also, results such as this one that show that even circuits with width 16 are extremely limited in depth (another relevant metric here is IBM's record $\log_2(QV)=7$ as of today).
Given this, what kind of circuits/applications can be executed with today's error rates that would require the full > 100 qubit processor?