This is basically why Google could be sued by Oracle, because Dalvik and their class library based on Apache Harmony were not a fork of OpenJDK.
Of course I cannot speak for Signal's protocol. Maybe it has no traps. I'm just commenting on that license. It's a strong license that makes some demands: good fit open source but bad for Google.
Which brings me back to the original question—why do we think that OWS's pushing of Signal Protocol is about money? Yes, I expect that for Allo they got paid by Google to write and maintain some code. But I don't think that their general claim "Signal Protocol is good crypto for everyone solving this problem" is motivated by money, because so many people solving this problem could use the GPLv3 version.
As Allo is not GPLv3, they obviously got it under another license.
(And it's not so obvious to me. The thing I linked is licensed under GPLv3 + MPL if used on the App Store. You can totally ship an Android app that runs a separate GPLv3 subprocess, and an iOS app that uses it under the terms of the MPL. The GPLv3-subprocess thing is what JuiceSSH does for running Mosh.)
What?
The protocol is publicly described. They've blogged about it. I can imagine people being able to reconstruct it from memory.
The first Google result for "signal protocol license" is https://whispersystems.org/blog/license-update/ , clarifying that it's under GPLv3 (i.e., patent grant) with an exception for the App Store. Has anyone paid money to license the protocol? Has Signal asked for money? Is it even possible to give them money for the protocol?