John Seeley Brown, in
The Social Life of Information, has a pair of anecdotes from Xerox's repair staff and either Xerox or Kodak (I think) customer service.
The support techs used to have a morning coffee session before heading out into the field on their calls. These turned out to be hugely useful tech transfer opportunities as repairmen would share both their unsolvable problems (machines which had broken and couldn't be fixed) and "This One Weird Trick" solutions.
The second anecdote involved customer service operators, who had previously, if I'm remembering this correctly, shared floor space with techs, but had been moved to their own space. Following the move, there were two specific operators who were able to help out customers by not following the specific scripts they'd been provided, but rather by supplying actual useful questions, diagnostics, and suggestions. When studied by management, it it turned out that one was a veteran of the former seating arrangement. The other, however, was a more recent hire. She did, however, sit next to the more senior rep, and had picked up tips listening and talking to the other employee.
The lesson I'm drawing is that you don't need people to be interrupted all the time. But you do need them to have some opportunities for a free-wheeling exchange. Finding the right balance point between heads-down work (the Xerox techs, for example, were mostly on the road and at customer sites, so they wouldn't have had input from other techs), and exhange opportunities.
In academia, there's the teaching vs. studying/research balance. A few days of lecture and sessions, with other time devoted to actual research, gives that balance of peturbing your own thinking and exposure to new ideas, and focusing on your own work. E.O. Wilson addresses that in his recent book, Letters to a Young Scientist (along with advice to avoid as much teaching and administrative responsibilities as possible).