There are two critical aspects to the bus routing problem. One is that no matter how well you design your system, there is always variance in the arrival times of a bus at any given stop. If you expect people to switch buses, then you need to account for this variance, and this means adding buffers. Nothing makes people stop using buses faster than missing your connection because your bus was late.
The other aspect is the what city topology you are dealing with. In square grid cities, you can probably put a tram on every road, and with one switch over, get to where ever you need to get to.
But many organically grown cities end up using the hub-and-spoke model, where there are main stations where many different buses meet. People switch over to the next connection (and you need a buffer here). Critically, you need all the buses to meet at roughly the same clock time, say every 30 min. Now, one thing you realize immediately is that not all routes are equal. One route might be only 25 min, Either you make it longer and waste fuel, and time for everyone sitting on the bus, or you wait an extra 5 min at the main station.
Bus scheduling is very difficult problem in real cities with weird topologies and real traffic issues. Buffers are a necessary part of any reasonable solution.
I'm coming from a different perspective: regardless of all else (all those issues you raised are very real), people need to get where they are going in a reasonable amount of time. Most bus service fails to account for that, but if you can't get people there in a reasonable amount of time there is no point in trying.
I agree that a system that does not deliver is going to fail. Transit systems can have improved scheduling in two ways:
(1) Better scheduling system. My opinion is that most real world systems are not too far away from the optimal trade-off curves. There is always room for improvements, or choosing better trade-offs, but it will rarely drastically improve things.
(2) More ridership: Most problems with speed just disappear if more people ride. For example, you do need a solid buffer when buses come every 30 min or more. But buses that come every 10 min or less, you can get rid of all buffer. A lot of scheduling problems are just not-enough-users problem.
No transit system has the money needed to provide great transit to their entire city even though if they did it would save all residents a large pile of money. They compromise on only the densist areas which are easy but mean that you can't get anywhere else reasonably. Of course some areas will always be easier than others to get to, but far too much of any city is not reasonably reachable without a car and that is a problem.
The other aspect is the what city topology you are dealing with. In square grid cities, you can probably put a tram on every road, and with one switch over, get to where ever you need to get to.
But many organically grown cities end up using the hub-and-spoke model, where there are main stations where many different buses meet. People switch over to the next connection (and you need a buffer here). Critically, you need all the buses to meet at roughly the same clock time, say every 30 min. Now, one thing you realize immediately is that not all routes are equal. One route might be only 25 min, Either you make it longer and waste fuel, and time for everyone sitting on the bus, or you wait an extra 5 min at the main station.
Bus scheduling is very difficult problem in real cities with weird topologies and real traffic issues. Buffers are a necessary part of any reasonable solution.