repeated 'sort by PID' doesn't reverse ordering for equivalent sequences

For this alignment:

>a
aaaaa
>b
aaaac
>c
aaaac
>d
aaaac
>e
aaaac
>f
cccca

1. Sort by PID yields a,c,b,e,d,f (decreasing on PID to a)
2. A second 'Sort by PID' yields a,b,c,d,e,f. (decreasing on PID to a)

Ideally, the first sort should sort but preserve ordering of alignment (order unchanged in this case)
The second sort should *reverse* ordering (order reversed for ties and sort metric).
The third sort should recover the first sort ordering.