It would be interesting to check whether that particular process has started to page-fault into the virtual memory file. The behavior you describe is typical of programs that reach a system limit on how much physical memory a process may consume before it starts going virtual. Doing so means that you reached some kind of threshold.
If you are on Win10, you can TRY this: While running (or preparing to run) your activity:
Open task manager
Select Performance tab
At bottom of that screen, click the hyperlink to Open Resource Manager
Select the Memory option
The table will list biggest memory users first. The column that says "Working Set" tells you how much physical memory some task is using. When a particular process starts going "virtual" (using the advanced virtual memory features of Windows), the Working Set will "plateau" and stop increasing. Unfortunately, another mechanism, much simpler, could ALSO be in force.
Depending on how it was written, Adobe's code could (and probably does) have buffer limits. If the PDF you are searching is bigger than the buffer limits, then Adobe's code would have to do the same thing as a virtual demand-page algorithm... start to load things from the disk because it doesn't have enough memory to hold everything at once.
Either way, it sounds like you reached the point where splitting might be a winning strategy. And in the long run, it doesn't matter whether it was Windows and virtual memory paging or Adobe and buffer sizing limits that you tripped over.
