Caerbhaighlien sources are not controllable, therefore they either replace usable source holdings with unusable ones or effectively reduce the source potential of a province. The BoM deigned to mention which interpretation was correct. The BRCS team appear to have figured that something described as rare and wonderful should be positive, figured that a plot device McGuffin (we need to access the ancient caerbhaighlien to have the power to defend our province) was better than a sore point (ok GM, how do I get rid of the Caerbagh crap polluting my source?) and so decided that the Caerbhaighlien sources would be additional to the normal sources of the province, not instead of them.
Leylink has similar issues; it was out of sync with the treatment of other domain types and also become obsolete with the concept of control by a council of scions rather than a single scion - the BRCS therefore removed the existing 2e anomaly and fixed it with a more versatile generic mechanic.
Part of the 2e to 3e conversion was to remove randomness from holding income thus technically all the old income rules were ignored… The ‘high level sources act as virtual guild rule’ was retained as an optional rule in BRCS and is used in some PBEMs.
In general the BRCS appears to have tried to follow comments in all products as far as possible, I have to admit I’ve not been aware of anyone splitting the ‘accessory’ and ‘expansion’ concept in the way you have, to me the terms are interchangeable, but everyone will rank products in terms of importance based on a number of factors, this can put RoE on top (prime source) or the bottom (first attempt corrected in later books) dependig on point of view. Personally I go for whatever works best and is most consistent in my campaign.
My understanding is that flavour text will convert either to explain issues which came up in play with more clarity, reflect the fact that the typist gets bored just copying, or be an attempt to avoid copyright issues. Rules issues will tend to change to improve consistency across the ruleset, fit with the 3e philosophy better (i.e. the change from random income to fixed), update earlier ‘canon’ for later ‘canon’, or because the players on the forum at the time simply preferred the new rule in gameplay.
Alignment as a ruleset has changed frequently from system to system and that does need to be reflected in the conversion; to be honest though alignment is so poor a system given modern personality tools that I find it meaningless - I’ve had characters described as CE by half the group and LG by the other half based on their interpretations of the characters actions - the 4e version which assumes that chaos is bad and law good is even more broken than the 2 and 3e ‘balance’ alignments imho. This is one sacred cow which should have been slaughtered early in the 4e process…
Part of the 4e conversion will be not only considering the relative importance of sources, but also identifying ‘canon’ that is inconsistent with 4e philosophy. My current worry is that BR was aimed at the gamers who wanted the ‘outside combat’ stuff and ‘role not roll’ play, whereas the 4e conversion seems to have dumped any consideration of either game style. I have no problem with a game that consists wholly of PC creation, shopping, and random acts of violence, but I want more from BR and a conversion may wind up having to ignore significant chunks of 4e design philosophy.