Try to address flaky tests in limit_change

src/qualitative_networks.jl

@@ -307,8 +307,12 @@ end
 
 """
     $(TYPEDSIGNATURES)
+
+Returns the limited value of `next_value` which is at most 1 different than `prev_value`.
+
+It is also never negative, or larger than `N`.
 """
-function limit_change(prev_value, next_value, N::Integer)
+function limit_change(prev_value::Integer, next_value::Integer, N::Integer)
     if next_value > prev_value
         limited_value = min(prev_value + 1, N)
     elseif next_value < prev_value

test/test-qn.jl

@@ -68,11 +68,14 @@ end
 end
 
 @testset "Async QN" begin
-    for i = 1:100
-        async_qn = aqn(network, N + i)
-        step!(async_qn, 10)
-        @test all(get_state(async_qn.model) .<= N + i)
+    for N = 2:5 # a few different levels of N
+        for _ = 1:100 # 100 different initializations
+            async_qn = aqn(network, N)
+            step!(async_qn, 100)
+            @test all(get_state(async_qn.model) .<= N)
+        end
     end
+
 end
 
 @testset "Get attractors" begin