Reference solution is definitely wrong:

['IjOIj;T"}Ug', '}1xN>Ky?99G[$ 3q3(kru=>0 ', 'IjOIj;T"}Ug', 'IjOIj;T"}Ug', 'IjOIj;T"}Ug', 'IjOIj;T"}Ug', 'IjOIj;T"}Ug']

'IjOIj;T"}Ug' should equal 'IjOIj;T"}Ug}1xN>Ky?99G[$ 3q3(kru=>0 '

The expected value doesn't even loop in the RDS text, so it couldn't be the correct value.

Random tests sometimes do not input any text bits before requesting the concatenated string. This case has never been specified.

test.it("#7")
rds = RdsConcatenator()

rds.input_rds_text("Hello ")
rds.input_rds_text("world!")
rds.input_rds_text("Hello ")
rds.input_rds_text("world!")

test.assert_equals(rds.get_text(), "Hello world!")

rds.input_rds_text("Hello ")
rds.input_rds_text("banana, ")
rds.input_rds_text("potato")

test.assert_equals(rds.get_text(), "banana, potato")

rds.input_rds_text("banana, ")
rds.input_rds_text("Hello ")
rds.input_rds_text("world!")

test.assert_equals(rds.get_text(), "Hello world!")

The last suite was not explained in the description

Other than that, it is not clear what you mean by this:

Remember, the strings will sometimes repeat, so you wrap have to wrap up somewhere (example 2 and 5).