>>Is it weird that some publishers see an obvious and substantial decline in EPCs and others see none at all? <<
Not at all, which is, I guess, your point! Fact is google can play around however it wants with payouts - for example reducing revenue share for certain keywords, - or even giving certain sites better shares so they stay - all is possible and sensible if they are trying to optimise the value of Adsense to advertisers and the network as a whole.
Ad inventory is another possibility. If a lot of webmasters are targeting the high value clicks, ad inventory will run out quick, and lower value clicks will take up the space.
Ive got no doubt at all that Google is experimenting on what sites not only provide the best CTR and RPC, but also the "added value" of individual sites or site "categories" to advertiser's poss ROI and conversions (as far as they can).
It would also make sense for them to encourage sites in niche areas for which there is an unmet inventory no matter the CPC, and which they cant meet by search site delivery alone.
Another thing they could look at is the number of other links on an Adsense page. If Adsense links are really the only option, it follows that the "value" of the click to the advertiser is not as high as if the reader had other informed choices. In that case they could reduce the delivery of certain ads to those pages and reduce the cost to the advertiser and revenue share to the publisher.
I know this all makes the system more complicated but thee are just examples of how Google could increase the effectiveness of Adsense, and maybe as a byproduct reduce revenue to certain pages/sites.
Im not saying at all that im 100% correct or indeed any of those reporting reductions in clicks have "less effective" sites.
All makes sense to me...