For two players sitting at different tables, there is NO covariance. Remember that covariance is a measure of the tendency of two hands to have the same result. The outcomes of two hands at separate tables are entirely independent events. In this situation, you can calculate the Kelly fraction, based on your combined bankroll, just as if you were playing alone. And you can calculate the resulting variance of a given number of total hands played in the same fashion as though you were playing them all yourself, at one table.
If you are playing two hands simultaneously at the same table, the problem is very different. The covariance between two hands differs from game to game. You need to find a source to look up the covariance for each game that you're interested in. Like variance, covariance will change a bit depending on the strategy that you're employing. The covariance figure for playing basic strategy may not be strictly accurate for the strategy you're playing. For MOST games, variance and covariance don't change a lot depending on play strategy, but there are exceptions in the case of some games if you're diverging wildly from basic strategy. The book to which you refer gives covariance for the variety of the game you are discussing, on the page you cited. If you don't have a source for covariance, it can, in principle, be calculated from the rules of the game, but for many or most games the problem is essentially intractable. Alternatively, if you have a facility to sim a large number of trials, you can get a close estimate of covariance that way.
At the same location, the book you mention gives a formula for calculating the resulting variance for each player. You can use this resulting variance in ordinary fashion to calculate the variance of a given number of total hands played. You can also use it to estimate the Kelly fraction. Remember, however, that the commonly-seen formula for estimating the Kelly fraction using, as the only variables, expectation and variance, is ONLY an estimate. It's a very good estimate for games like blackjack in which a hand usually results in a one unit win or a one unit loss. The estimate can be seriously inaccurate for highly-skewed (lottery-like) games, so be careful.
