Agreed on the difficulty: you would aim first at the low hanging fruit for placing the taps, the easiest being the onshoring spots. But... for the MEMS, I cons

Agreed on the difficulty: you would aim first at the low hanging fruit for placing the taps, the easiest being the onshoring spots. But... for the MEMS, I consider this as one way only (public, probably crude/basic) and very possibly we'll learn of other techniques that are usable in say, 5, 10 or 20 years. That's what our taxes are financing in the end ;-) research on fundamental new ways to tap the neighbour/hostile country's fiber. Lastly, I wouldn't sweat it too much on data/speed capacity but more on power needed to process and Y-split this datafeed to exfil. If in the future we got to learn that micro/nano-nuclear powerstations were tied to submarine interception points, I wouldn't be surprised at all. The data/speed capacity is probably the less difficult part compared to seamless insertion, power and exfil system. Now all these are just guesses based on what's public so far thanks to Snowden & al.