diff --git a/src/MacroModelling.jl b/src/MacroModelling.jl
index 25b43117..5130ca56 100644
--- a/src/MacroModelling.jl
+++ b/src/MacroModelling.jl
@@ -3217,7 +3217,7 @@ function solve_steady_state!(𝓂::ℳ, symbolic_SS, Symbolics::symbolics; verbo
                     # TODO: try have this run for 1000 and and use closest_solution based on the previous result and not on the cache. that way you dont crowd out good and diverse solutions in the cache and make sure he finds the other SS. rely on previous commit for way of implementing closest_solution
                     while range_iters <= (cold_start ? 1 : 500) && !(solution_error < 1e-12 && solved_scale == 1)
                         range_iters += 1
-                        fail_fast_solvers_only = scale == 1.0 ? false : true
+                        fail_fast_solvers_only = range_iters > 1 ? true : false
                         # println(range_iters)
                         # println(scale)
                         # println(solved_scale)
@@ -3976,7 +3976,7 @@ function block_solver(parameters_and_solved_vars::Vector{Float64},
             end
         end
     else !cold_start
-        for p in (fail_fast_solvers_only ? [parameters[1]] : unique(parameters)) #[1:3] # take unique because some parameters might appear more than once
+        for p in (fail_fast_solvers_only ? [parameters[end]] : unique(parameters)) #[1:3] # take unique because some parameters might appear more than once
             for s in (fail_fast_solvers_only ? [false] : Any[false,p.starting_value, 1.206, 1.5, 0.7688, 2.0, 0.897]) #, .9, .75, 1.5, -.5, 2, .25] # try first the guess and then different starting values
                 # for ext in [false, true] # try first the system where only values can vary, next try the system where values and parameters can vary
                 for algo in [newton, levenberg_marquardt]