Erik,

Don't get me wrong.
I agreed using one memory byte as per your example. That is why I advice Nutty Blue in my previous ad, like a good solution for the problem.
The Timer2 driven by external square wave signal it's just another way to solve the problem and extend the 64ms limit. However, I said it's not common practice. There are a lot of circuits with few counters stage and oscillator inside one unit easy to adapt in order to get 3Hz square wave signal. But only if will accept additional hardware.

You wrote:
3) How is enabling a square wave out from T2 (C9h enables this) going to change the timers operation?

Bit C9H is C/T2 mode operation. Setting this bit will switch the timer2 clock input from internal to external pulses coming from T2 pin. I never said that setting bit C9H enable a square wave out from T2. In fact to enable a 50% duty cycle clock to come out on P1.0, the Bit C9H must be cleared and ORL T2MOD, #02H.

Kind regards,
Silviu