Voting protocol - How many dishonest tallying centres can the protocol cope with?

We have the voting protocol that is described here.

I am looking at the following exercise:

Determine how many dishonest tallying centres the protocol can cope with before the following properties no longer hold.

1. Only authorized voters will be able to vote.
2. No one will be able to vote more than once.
3. No stakeholder will be able to determine how someone else has voted.
4. No one can duplicate someone else's vote.
5. The final result will be correctly computed.
6. All stakeholders will be able to verify that the result was computed correctly.
7. The protocal will work even in the presence of some bad parties.

Could you give me some hints how we can determine that??

• Addition: the original source of the protocol is Nigel Smart's freely available Cryptography, An Introduction, with the protocol here. $\;$ Hint: the protocol uses Shamir's Secret Sharing in $t$-out-of-$n$ mode, which ensures that the correct tally can be obtained from the data published by any $t$ correctly-behaving tallying centers, where $t$ is a parameter less than the number $n$ of tallying centers. – fgrieu Jun 18 '15 at 14:04