Sir TLUL
Comments
It's worth noting that the only two other bug reports showing up for a search for root.tick.level.connection are MC-11571, which is about a one-time issue on block breaking rather than a continuous lag problem, and MC-12799, which does not provide enough information to conclude whether or not it is the same issue as this or not. I rather think that it isn't, because this particular bug occurs only along a fairly narrow "corridor" of sorts, rather than in a larger general region as MC-12799 describes.
Wouldn't the better fix be to pipe all isDown queries through a decoder method that handles -100, etc as the correct mouse buttons? This would also mean future input methods could be more easily integrated with legacy code. This isn't a frequently-called method (relatively speaking) so it would add very little overhead.
Still present in 1.7.5, 1.7.6-pre2, and 14w11b.