It had to have been, in order to implement authentication and password reset. If not, Skype has been sitting on a massive breakthrough :). Decentralized user/password auth for 100M users would require a few GB of storage. Then you'd need a way to verify that info. Not impossible, but see how heavy e.g. the Bitcoin block chain is. The real obvious tell is that Skype had password reset via email, so obviously they've always had a "backdoor".
On top of that, the client was heavily obfuscated, so it's unlikely anyone besides the developers actually knew what it did in every situation. (Skype client did stuff like generate executable (.com) files on the fly to check things - they really didn't want the protocol reverse engineered.)
Skype allowed LE in well before Microsoft. And how would that even be noticed? Your system would relay through "random" peers and you had no way to verify the selection, and since the protocol is unknown, you have no way to verify the packets transmitted. The client could easily divert your call via a special interception node and you'd have no idea.
Skype could have easily implemented proper E2E crypto with a simple audio verification step (ala PGPFone and now ZRTP.) They chose (and continue to choose) not to. They could then publish the spec so users could verify; they choose not to.
What you're probably thinking of is MS's decision to nix the P2P part of the Skype client which imposed undue resource usage on clients. Now MS runs the "peers" needed to maintain the network. This is the least concerning part of Skype, but looks the most sinister to a casual glance.
