I read the source code, and here are 2 improvements:

1) in get_block, we can replace "dec count" with "cpy count" It should save one or 2 cycles, if I'm not mistaken

2) get_countdown can be rewritten as follows:

        ldy #$89
 c4:
        cpy #$89
        bne c1
 c0:
        jsr get_byte
 c1:
        ldy #$89
        bne c2
 cx:
        jsr get_byte
 c2:
        sty tmp_cntdwn
        cmp tmp_cntdwn
        bne c4
        dey
        cpy #$80
        bne cx
        rts

which saves probably 4 bytes