You can chain handlers by simply rewriting the interrupt vector. Something like this (for a sequence of two handlers):
handler1
; ...
lda #<handler2
sta $fffe
lda #>handler2
sta $ffff
lda #line2
sta $d012
rti
handler2
; ...
lda #<handler1
sta $fffe
lda #>handler1
sta $ffff
lda #line1
sta $d012
rti
To my intuition it seems it would be quicker, though slightly larger.I've only skimmed through the article so maybe there's a reason the author considered this and opted not to. Running out of memory maybe?
Nitpick: I think he still has the Kernal banked in, so he'll need to use the soft vector at $0314. But that is exactly how I've done chained raster splits otherwise. It's really the most straightforward approach.
If you put both handlers in the same 256 byte page you can get away with only updating the low byte of the address.