When one side of an entangled pair is observed [...]
Whether and in which way a pair had been entangled can of course only be known from both having been observed, and both observations having been (separately) evaluated and the results having been correlated.
Is it possible for that side to detect or infer whether or not the other side was already observed at that point?
Considering only one side of a pair there cannot be made any inferences about "the other"; it cannot even be inferred whether or not some particular "other side" might have been "in play" at all.
But it is certainly possible for one side to detect the other side having broadcast their results, before collecting observations pertaining to the own ("one") side of the pair. If so, the two events to which the observations belonged were time-like related to each other.