Thanks for sheding some light here!
I am sure you are technically and factually correct.

Often though such knowledge does not manifest itself the way your post implies ("it should just work - same network same base station") to users that suffer from connectivity issues and can't get those resolved easily. So some question still remain.

If one has frequent connectivity issues - would that not impact "sufficient data in the buffers" on both ends? They just know that even though they have asked for a huge chunk of data to arrive it is just not comming to their n900.

If there is sufficient data in the buffers on the other hand - would modem in 3.5 not need more power to receive/transmit?
What happens if it does not get that power quickly enough?

I lived through it (including frequent crashes while attempting to use MobileHotspot) - so I can empathise with those that experience thouse issues. They are real, and reflashing alone did not help in my case.

The only thing I saw was an icon going from 3G to 3.5G and bars for signal strength flickering like a set of christams lights, no dat was arriving and no meaningful help from anyone was forthcomming.

I think that understanding of what is going on is very important, but that knowledge itself will not solve the practical end-user impact.

If you put your phone in super starving mode on overclocked kernel you might be able to see how 3.5 mode shuts down your phone which was working fine in 3G mode.

Maybe the power management of the n900 is not anticipating/not aware of "sufficient data in buffers" causing some instability at times. Or maybe ppp based 3G modem connections with no wlan0 interface in the picture are more resiliant. Or maybe another APN, or no 2G switching, etc...

I might be way off base with my observations, but...
I had those issues and although I believe that they are mostly network operator specific - what little I have done on my n900 has seemed to have had helped me.

Anybody's else milage might vary.

netC