I know of some pools who are trying to implement merged mining. However ATM there are some perfomance issues that even MasterPool suffers from at about 50GHash/s for which I have workarounds in place but it needs to be fixed permanently. Not to mention that MasterPool architecture was designed to deal with this splitted blockchains stuff from the beginning. Everybody is working right now lowering the foot print of merged mining. You have to understand that big pools optimized their complete architecture for mining one blockchain. Now they have an additional blockchain, merged mine proxy and additional calculations in their backend. It feels like hitting a concrete wall with a jet

The problem is you CAN'T simulate large load easily on testnet. Thus I guess some pools are turning it on occasionally in order to make sure merged mining optimizations are working properly turning it off if they gathered enough data to do the next optimization iteration. So please don't panic and wait a few days.
Nodemaster pointed that out very precisely. I want to dispel fears of any of you who's thinking that some big player is "stealing" coins for himself.
Yes, I'm testing merged mining on my pool with some backends. And no, I'm not going to "steal" those blocks for myself. But implementing MM on pretty big pool is really pain and I didn't want to announce it before I'll be sure it will work *somehow*. But after spending ages on weird bitcoind crashes using MM together with my custom performance patches, I'm starting to be pretty optimistic.
I need to finish GUI for NMC support, then I'll also start giving away all those mined blocks (already over 100 NMC block mined during my tests) back to pool users to spread Namecoins between people.
Btw first merged mined block ever was #148744 (BTC) == #19274 (NMC)