Also, the implementation as it stands is in any case flawed, since specifying both timeout and endtime is allowed by the code, and results in endtime overriding timeout silently, and in the posix version the resulting timeout error message will have the ignored timeout in it.

Unfortunately this has been in here for a while (March 2011) so people may be using it.

Based on the docs in the original commit (c4a0fa6e687c), I suspect endtime on wait was supposed to be an internal parameter....except that it is never used internally.  The endtime is converted back to a timeout in order to call wait.  Ideally we'd just drop it...but it may be in use in wild.