Excellent question!

First we allocate 100% of traffic to the chosen variation by always returning true whenever the track_experiment method is called.

if et.track_experiment(“exp_name”): // <- always true
   self.run_new_feature()
else:
   self.run_old_feature() // <- this will never be executed

Then we clean up the redundant wrapper from the code.

self.run_new_feature()

Since we are already allocating 100% of traffic to this code path, rolling out this change should be safe, because it has no impact on production behaviour.