If a light splitter and combiner introduce a 90o phase difference in the output beam(s),
In the last two 'blocked' cases, beams at
0o and
180o, and beams at
90o and
270o do not interfere and cancel as they are on spilt paths. Beams at
180o and
270o phases are visible beyond the outputs of the splitter. These outputs are at opposite phase to the other outputs of the splitter, but do not interfere and cancel with them because they are on different paths.
Not weird.