Does it become problematic to have more observables (e.g. four observables) when it comes to estimation? What I meant is having more observables lead to less chance of convergence?
No, ideally more information will lead to a more concentrated likelihood and easier estimation.