Both are FRED series (CBBTCUSD and WALCL). Rolling one year correlations of one week deltas:

    0.3 +-+----+------+------+------+------+-----+------+------+------+----+-+   
          +      +      +      +      +      A     +      +      +      +      +   
          |                                 A                            AA    |   
      0.2 +-+                                 AA                        AAAAAAA+   
          |                         AAA  AA AAA                        AA      |   
      0.1 +-+                   AAAAA AA  AA    A                     A      +-+   
          |                   AAAAA   AAA         A                  AA        |   
          |                    AA                A                             |   
        0 +-+                 A                 AA AA AAA                    +-+   
          |                                      A   AA A                      |   
          |  A                                      AA   AAA        A          |   
     -0.1 +A+       AAA     AA                             A        A        +-+   
          |AAA      A  AAAAA                               AAAA                |   
          |A  A           A                                                    |   
     -0.2 A-+ AAAAA        A                                  AAA AA         +-+   
          A     AA AA                                          AAAAA           |   
     -0.3 +-+     AA                                                         +-+   
          |                                                         A          |   
          +      +      +      +      +      +     +      +      +      +      +   
     -0.4 +-+----+------+------+------+------+-----+------+------+------+----+-+   
          2017/1/1                      2018/11/25                     2020/10/11

So ELI5 what I am seeing here. Aren't those correlations all over the place?

Yup, that's my takeaway. You see the same thing when you use future Fed moves as well (which is one way of checking that this graph isn't missing out on some announcement effect, as assets lag announcements).