The key point in DPOS consensus mechanism is how to elect the representatives(or witness) producing the blocks. You described the mechanism as "open community poll", but how to organize the poll can result in a very different governing structure.

For example, Steem has 20 witnesses plus 1 time-sharing witness, and each share can vote 30 times on different witnesses. The interesting part is why each share should have 30 times voting chances.

Under this rule, to become a witness does not necessarily depend on the majority of "community support". Instead, cross-voting structure among top 10 or 20 whales becomes the most important decision factor. If you get 20% voting, it is pretty safe to maintain the witness position and to get that amount of voting, you need just 10 other whales supports with at least 2% of the total Steem shares. You don't need to worry about all other community members' voting at all. So for the top whales, it is very "reasonable" to make agreements to do cross-voting each other and not to "betray" each other. There may not be a just one power block, there can be two or three different power groups voting internal members mostly. If one member betrays, it is very easy to pull down his seat.

If the system allows only one voting chance per share, this kind of cross-voting is impossible. Maybe one big shareholder can divide up his shares to multiple accounts and can occupy multiple witness seats, but it is very difficult to hold more than several unless he has more than 50% of the total shares.

When the number of representatives(or witness) is small (such as less than 100), and having multiple voting chances per share, it is easy for the top whales to collude to make a power block whose main goal is to maintain their prestige (and very profitable) positions.

This kind of DPOS poll system may be okay for a consortium chain with a limited number of established pre-existing power players such as a big enterprise consortium network, but not idealistic for a public chain, in particular for a public contents network where those witnesses can decide which content is good or bad.
Regarding this issue, I wrote a post several months ago,
https://steemit.com/steemit/@atomrigs/net-neutrality-and-curation-rewards

If I have a chance to design a new Steem system, I would do these things;

• At least 100 full-time witness nodes
• Only one witness vote chance per share
• Witness should have no curation voting power
• Witness rewards in terms of dollar amount should be capped.