I can answer some of these:

1. Historical performance does not imply future performance, so don't spend/invest more than you can afford to lose. (But the algos can be lucrative if the conditions are right.)

5. You should expect many of your strategies to become useless after some period of time due to changes in the market. Prepare to be very flexible. Objectively analyze your algos and retire the poor performers.

Bonus: This is easy to do if you have a dedicated VPS and a separate development machine.

EDIT: Where did the bonus question go? Now this thread will be confusing to latecomers.