The movie named the Matrix made famous the concept that a deja vu event we encountere is proof of a glitch in the matrix, a sign that the masters of the illusion changed something. I myself also notice other proofs of changes in the matrix: quite often in my work as a software engineer I notice certain weird behavior in the program I am testing, but then when I try to replicate the error, it won't happen.
Over time I've learned to let go of situations like that if two or three retries don't make the perceived error surface. No point in pursuing what is not persistent in the simulation, it seems...