What you've experienced is very common. I doubt very much if it has anything to do with harmonics by which I'm assuming you mean barrel and action vibrations. They affect group size rather than MV spreads and determine which pressure / MV combinations / ranges are countered by barrel movements to give the same POI.
A 0.2gn charge weight change is very small in relation to c. 46gn, and there are several factors at work. First, scale accuracy as beam scales and most handloaders' electronic scales are only advertised as being accurate to plus or minus 0.1gn. So if you're at the top end of the 46.2gn tolerance, and the bottom end of 46.4gn, you've loaded two lots with 46.3gn. Moreover, unless you keep your head and eye position in relation to beam scales absolutely constant by referencing the view in some way at the start of the session and ensuring it is repeated each time you look at the beam position, you can add another 0.1gn or more variation into your loads causing the potential variance to exceed your planned weight change.
Then too, smallarms cartridges often produce much larger pressure variations than you'd expect. The only manual I've seen that show this is Any Shot You Want, the A-Square reloading manual. Here's what it gives for .308 Winchester, 180gn Nosler Ballistic Tip and IMR-4064 in CUP units:
41.0gn: ..... avg 42,500 CUP ...... SD 4,700 ....... ES 10,100
44.0gn: ..... avg 55,800 CUP ...... SD 2,800 ....... ES 7,100
45.0gn: ..... avg 59,900 CUP ...... SD 1,800 ....... ES 4,100
With so much potential variation in strings of test shots, a mere 0.2gn charge increase over 46.2gn, only 0.43%, can be lost in the statistical noise of other such factors. Note the starting load and its relatively low pressures produces the largest variations which reduce as the cartridge and powder achieve full working pressures. This is a 'bad example' but every load shown in the manual sees some significant pressure variation.
Dr Geoff Kolbe founder and proprietor of Border Barrels, former experimental research physicist and an internal ballistics expert also describes rifles and their cartridges as suffering from 'non-linear effects'. Unlike large calibre cannon and artillery where designers accurately computer-model the effects of individual and combined changes such as projectile weight and shape, propellant characteristics, barrel and rifling design etc to produce a compromise that meets the desired ballistic objectives, Dr Kolbe says that a small change in one aspect of rifle cartridge design can upset the applecart on all the rest. Also a change to one aspect like propellant charge weight on its own may or may not provide linear results, and may even produce them up to point X, but with a different result's graph line following from there when that point is passed. In rifle cartridge handloading terms, your results might not be duplicated with a different bullet or with the same bullet at a different seating depth and so on!
The science of smallarms cartridges powder charge ignition and burn is shall we say 'an inexact one'. Thanks to the BR and PPC, many people believe as an near act of faith the short, fat, sharp shoulder case shape gives greater pressure and ballistics consistency than those with same capacity but long thin cases and shallow shoulders. Yet the .300 H&H Magnum which has everything wrong about its shape gives as good, often better, ES and SD values than the identical capacity, identical SAAMI PMax .300WSM with its much 'better' shape when the same load combinations are used. It may be that whatever cartridge and bullet combination you're loading is marginally unhappy with 46.4gn for reasons that I doubt we'll ever know - or you're getting this result twice is simply a statistical quirk. Load up and chronograph enough cartridges and you see almost 'everything' sooner or later!
Finally, you don't quote the string size or the velocity variations around the average. In small test strings of 3 or even 5 shots, one that produces a small overall spread produces a much more useful arithmetical mean than a 'neighbour' with a larger spread.
eg let's say you fired 5 test shots at 46.2gn and every one produced 3,150 fps. The cartridge didn't like 46.4gn and/or the potential range of statistical spreads kicks in and your 5 shots run at 3,120, 3125, 3,155, 3,160 and 3,170 fps, the average drops below 46.2gn's result to 3,146 fps. Unlikely to get such a difference in 0.2gn and to get it twice, but still possible.
The overall message is not to bother about quirks like this - they happen. On the other hand, if 46.4gn gives consistently smaller or larger groups than 46.2 and 46.6gn, that is telling you something useful.
