Not my Favorite one.
Bonus didn't make much since to me.
In fact, I find the answer to the bonus question a bit misleading. The outcome is the average within the grouped data (by year, month, day of the week, hour) which favours data sets with one strong year.... As is the case for 2 pm on Sundays in August, which is sampled only for one year. There are 18 combinations of month, day of the week, and hour in the data which show a larger total number of lifts than 2 pm on Sundays in August (see screenshot below).
Since three full years are sampled, I would rather use the total number of lifts as an indicator as when to head for Tower Bridge, or, something like an average, wheighted by the number of the day-of-the-week being sampled in a given month (i.e. the numbers of Sundays and Mondays can be different within a given month).