With regard to the patents, he has nothing to negotiate. However, if he wrote the code, and released it under the GPL, he could definitely claim that their software which uses the technology is a violation of the terms of the GPL license. This claim might be legitimate (if they copied) or completely not legitimate (if they didn't re-use any of the code, and merely developed new software with similar capabilities and features).
In reality, OP is going nowhere with this. The USPTO is unlikely to invalidate the patent because somebody presents them with prior art that they already considered. Even if the USPTO was wrong, they aren't going to reconsider. Even if this was new prior art that nobody had previously known about... the USPTO is extremely unlikely to reconsider the patent grant. If it has been granted already, then, as they say, "prosecution on the merits is closed." Oh, sure, if IBM sues somebody, the prior art might be helpful. Claim 1 might or might not hold up unmodified in court, depending. Were it to require modification, then IBM could potentially lose a suit. IBM is unlikely to pursue the matter at this stage anyway, and they probably don't even realize that they have this particular patent.
I'm not going to read the entire document to make sure that I understand precisely all of IBM's definitions of terms, so I can't give an informed opinion on the validity. Without doing enough reading to be informed ('cause hey, this is the internet after all, I'm supposed to be wrong or at least uninformed), the whole thing probably turns on the feature of not having to re-link or even re-load the application to set the heap checking feature on and off.
Remember, you're only infringing a claim if you are infringing the entire claim. This patent isn't on "heap checking" or even "run-time detection of invalid heap access", or anything of the sort. If you aren't infringing the entirety of claim 1, the entirety of claim 12, or the entirety of claim 16 (the 3 independent claims), then you aren't infringing this patent. Similarly, if the HeapCheck software didn't do everything listed in those claims, then the differences between HeapCheck and those claims is the part that is (supposedly) novel and non-obvious.
(standard disclaimers apply; I'm no lawyer, this isn't legal advice, I'm just an engineer that happens to work the IP/patent process for my department)