Two trends motivate the following work. First, as the number of routing layers increases, the cost (via parasitics, via reliability, via blockage effects) of buffer insertion also increases. Second, as the integration of hard IP or noise-sensitive blocks in the SOC context increases, it is again more difficult to insert buffers into global interconnect on an as-needed basis. Our work investigates the problem of capacitive crosstalk, and considers the tradeoffs inherent in various "pure routing" solutions. One well-known example of such a solution is "swizzling" (i.e., permuting) parallel wires of a wide global bus so that no pair of wires will experience worst-case coupling for more than some fixed length.