The GPL doesn't necessarily care about the release state of the code. Rather, it cares about whether the code has been intentionally distributed or not. Hosting the code in a public repository on GitHub would certainly qualify as distributing the code.
As such, an errant pull request could have some onerous consequences to the copyright owners of the project. In theory, an application could be effectively forced to be re-licensed by pulling in GPL'd code.1
At the same time, the FSF has indicated a need for "intent" behind distributing code under the GPL. Since the inception of the GPL, various nefarious and hypothetical schemes have been concocted in order to force the release of source. However, the FSF has held that owners of the code needed to have knowingly added2 GPL'd code in order for the requirements of the license to take hold.
Said another way3 - nobody can force anyone to release their own source code through the GPL. The GPL is a license, not a contract.
So an errant git pull request wouldn't necessarily force an application to become released as GPL code. The original application owner(s) didn't intend to distribute and license the code that way. However, once the original application owner(s) is / are notified that their application contains GPL'd code, they are obligated to either remove the GPL'd code or re-license their application under the GPL.
1 It's also worth noting that there are ways to incorporate GPL'd code into an application without impacting the existing license for the application.
2 That's "knowingly added" or "reasonably ought to have known." The excuse of "Oh, I didn't know that GNU library was GPL'd!" is an excuse almost as old as the GPL itself. Repository owners ought to be checking the code that's being added to their application.
3 Hat tip to Robert Harvey for that prhasing