I'm happy to clarify as best as I can. Right now we are not charging at all for bandwidth overages. This will change at some point in the future but only after all users are notified well in advance.
Currently droplet bandwidth is essentially unmetered.
For your first question, the answer is yes. Transfer between droplets in different data centers currently must traverse the public internet so it will be considered in the bandwidth use.
On the 2nd question, there has not been a final decision made on how and if bandwidth will be pooled between droplets or how limits will work for droplets that exist for less than a month. These are things that will be addressed and explained in detail when users are notified of an upcoming change to our policy about charging for overages.
There is no ETA that I am aware of for this change and while limits have been listed for droplets since the beginning just as they are now (3+ years) these limits have never been enforced.
The exception to this of course is where massive bandwidth use is considered an abuse issue. Since hypervisors host multiple droplets we do ask that users who have applications that require continuous or near continuous bandwidth use configure their droplets to limit their use to 300Mbps.