If I were going to do this (and might, with my spare K4 frame, for experimental reasons, before carving into my beloved K2), I would start with the SERIOUS inspection of the existing frame tubes. They often rust, especially down below. Then, I would consider cutting out most of the 3-tube run from seat-to-head, followed by using either slightly larger diameter tubing or slightly thicker (smaller I.D., same O.D.) tubing as stock, making 2 uprights (each side) of smaller width in between, replacing the one of stock design. Then, the 2 outer tubes would need a little gusseting at their landing points on the uprights, because they will be landing into lighter tubing, which will tend to crack the welds otherwise.
Honda's frame design was predicated on the Norton Featherbed frame, one of the most successful frame designs of all time. Oddly, it was first built of thick water pipe, by a couple of semi-liquored Brits during a cold winter, who wanted to race with Norton's tremendous torque. The advantage is the 3-point head landing, coupled with the wide cradle, both of which tend to make a rigid triangle between the swingarm mount points and the steering tube, leaving the always-flexing engine out of the system. Honda's only mistake (starting with the K1 model) was to use thinner-walled tubing, which greatly reduced the side-to-side "twitch" reduction of the original Featherbed. Several racers I knew added a wrap of .031" tubing around the original side tubes on top (split them lengthwise, rewelded them over the original ones), to find they had created a frame so stiff it barely resembled the original (in handling terms). I've long wanted to do this to mine, but it REALLY complicated the act of removing the engine: that last 1/32" is really needed sometimes.
Racers I knew who added the "kits" of the 1970s that allowed removal of the upper tubes for engine maintenance all abandoned them for evil high-speed handling issues. I suspect that this was due to the poor method of reconnection and the even-thnner tube that was sold with those "kits" back then. The removable sections had 2 bolts that passed through the match-drilled holes on either end, supposedly squishing down onto the stock tubes when tightened. Problem was, the stock tubes simply crush, leaving a poor connection, and, frankly, one bolt per end is simply not enough grip. Drag racers got away with it because they don't often take wide, sweeping turns on purpose.
Part 2 of "if I were going to do this, and might...", if I were doing it for in-frame engine servicing...: the insert sections must tie in with very thick-walled joints that will not crush, with 3-4 short, tough bolts at each junction that thread into the inserts, not through the tubes. Do not use solid slugs, they are weaker than tubes (that's just physics, boys, and ask the guys who extended forks that fateful way). The joint, put simply, must be as strong as a weld. Personally, I would also attempt to add as much width over the engine as is practical, by either offsetting the pipes or bending them outward a little, to improve the lateral stiffness to the head. This I would do directly over the engine area, narrowing back to the head. Look at the CBX (1000) or some crotch rocket tube frames to get the idea: it works. And, while cutting in pipes and the like, think about a small gusset to box in the "square" behind the engine, above the swingarm pivot, to steady up the head. That works, too.
If I were doing it for racing, I would do it to the downtubes and rear-of-engine cradle instead, adding an extra bent-tube crossmember across the bottom rear of the engine. It would be stronger, and the engine would be lowered to the ground, whole, for servicing. I've seen this done, and quite successfully, on the "F" model on the circuits of the late 1970s.
Like I've got the time, in between all these swingarm rebuilds...
