Ah I didn't realize it was going to be a hydraulic roller. That would change my vote. The most informative of your choices would then be ICL changes, in my opinion.
The LSA comparisons are always interesting. One thing that I don't recall seeing is an LSA comparison with the same ICL. There have been plenty of LSA tests, but they always decrease the ICL as they decrease the LSA. e.g. 114 LSA with a 110 ICL, 108 LSA with a 104 ICL. They usually then conclude that the tighter LSA made more low and mid range power without considering the change in ICL.
As stated in the above post, Wide vs. Narrow LSA changes are interesting and often debated, regardless of whether you move the ICL independently.
Another idea would be use two cams, but increase the static compression ratio with later intake closing point. The ICL could be the only variable or it could be coupled with LSA. It might be a half to three quarters of a point of compression. I remember one LSA test where the tighter LSA's made more power (especially torque), but they also kept decreasing the ICL and therefore the DCR. If they kept increasing the static compression ratio with the wider LSA's and later ICL's I think the wider LSA's might have faired better. That would obviously be hard to do on one engine, but with two......
pl
I won't be varying the LSA a lot (or probably not at all) because I need to watch overlap. These are street engines that need some vacuum. I look at LSA as an afterthought anyway as I center a lot of things around overlap. If you keep widening the LSA on an engine on an engine that needs overlap, then you're going to lose power. In the same token, if you keep tightening it up on an engine that can't handle it, then you push all your charge out the exhaust port....and lose power. I generally see a sizeable torque gain on specific engines with a tight LSA, to the point where I sometimes use a 107-108-109 LSA on a street engine used for towing or pulling, but the combination has to warrant it.
Two of the tests that I think would help guys the most would be the compression ratio test and the ICL test. I get a lot of guys ordering pistons and saying, "I want to get as close as I possibly can to X compression ratio because I want the most power I can get." If a half point of compression only gives you 10 hp, then would you risk a ping or detonation on 10 hp? I've seen trends on some engines where I'm comfortable enough to answer that, but rarely do we dyno two identical engines. Another one is, "Advance that cam as far as you can for low rpm power, or retard it as much as you can for high end power." It doesn't always work that way as it depends on the combination and how hard the cylinder is pulling on the port. I've seen combinations with really advanced cam timing that made more high end and average horsepower.
The last chance I had the opportunity to do this, I did two 445's, same heads, same intake, carb, compression, etc., but changed the duration split while holding the overlap and lift the same. I had a good hunch based on past dyno tests on what the difference would be, but it's good to see actual numbers. For any factory head, TFS, etc., I would never consider anything less than an 8-10 degree split based on the results of those tests. It was a 12 hp bump, which may not seem like much, but if you're paying for a camshaft, why not pay for 12 more hp?