In some cases, it may be wise to increase the tx fee (or increase one of the output sizes, if possible) instead of sending excess btc to a change address, as the cost of spending the input from that change address (in addition to the cost of storing and securing the private key to that change address, and the "cost" of potentially decreased privacy as a result of using that change address) may exceed the value of the input.

However 400actforsale's proposal is to implement a minimum output amount of 0.000546 btc, which is over 5x (what used to be) the standard minimum tx fee for transactions that do (didn't) qualify for 0 tx fees (to get confirmed by most pools). As a result of this, I believe the proposed minimum output is too large, however I might be able to support a higher minimum output size from what we have currently.

There are certain benefits, risks and costs to being in a situation in which you control the private keys of your bitcoins, verses storing your bitcoins at a service that has control over your private keys. One cost of storing, securing and backing up your private keys is that you are unable to invest your coins in the bankroll of a gambling site, or lend them on bitfinex, while a benefit is that the chances of your coins being stolen as part of a large hack/theft are much lower. When discussing faucet payout-type sized inputs, it will probably be advantageous to let a reputable service to maintain custody of your coins, as you will not need to deal with the high cost of spending many small inputs, especially if such service does not charge you a fee for holding your coins for you. (this ignores the fact that dealing with faucets is probably not a good use of your time, as you will end up making well under minimum wadge for even developing countries).