why is this VRP problem taking so long to solve with the AUTOMATIC first solution strategy? - or-tools

Question

I've modified the data in https://developers.google.com/optimization/routing/vrptw to my specific problem - see below.
If I'm trying with FirstSolutionStrategy.PATH_CHEAPEST_ARC or FirstSolutionStrategy.AUTOMATIC, it takes 42 seconds to solve this tiny problem!
If I'm changing to FirstSolutionStrategy.PATH_MOST_CONSTRAINED_ARC it solves within 3ms.
Why is the solver so sensitive to this setting? Is there some mode in which the solver tries all strategies in parallel and selects the fastest or the best one?
def create_data_model():
"""Stores the data for the problem."""
data = {}
data['time_matrix'] = [
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[1000, 99999, 99999, 99999, 99999, 99999, 73, 104, 113, 117, 99999],
[1000, 99999, 99999, 99999, 99999, 99999, 84, 99999, 122, 125, 99999],
[1000, 99999, 99999, 99999, 99999, 99999, 97, 75, 122, 125, 99999],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 61, 109, 114, 103],
[1000, 99999, 99999, 99999, 99999, 99999, 47, 99999, 83, 89, 77],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99, 92],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 88, 94],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 42, 64],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 51],
[1000, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999, 99999]]
data['time_windows'] = [(0, 10000), (435, 450), (450, 465), (450, 465), (475, 490), (480, 495), (540, 555), (540, 555), (600, 615), (720, 735), (810, 825)]
data['num_vehicles'] = 5
data['depot'] = 0
return data
also changed the time dimension:
routing.AddDimension(
transit_callback_index,
10000, # allow waiting time
10000, # maximum time per vehicle
True, # Don't force start cumul to zero.
time)