On proper Stack Exchange sites, poll-style questions are not considered acceptable. They typically would serve only to measure users' opinions in aggregate, which is not constructive.
However, meta sites are a bit different. Determining users' opinions is one of their purposes. The scores of posts representing different positions are often taken as a representation of community opinion. In many cases these are long, detailed posts, which provide one user's justification for a position, but in some cases they are short posts which merely express the opinion itself. In these cases, the answers are serving as a de-facto poll, even if they may not be labeled as such.
On Meta Stack Overflow we recently had a discussion about what tag should be used for the new Swift programming language. A few different opinions were expressed in answers and comments. However, it was difficult to get an exact sense for how much community support there was for each possibility: some answers suggested multiple possible tag names, and there were many comments and answers suggesting the same tag names.
In order to more clearly judge the community opinion, I created a follow-up question which was explicitly structured as a poll. A single answer was posted for each possible tag name, and users were instructed to vote for the ones they preferred. If users wanted to respond at-length with an opinion that would not fit into the comments on the new question (none did), they were instructed to post answers on the original question. I wanted to keep the the limited to the poll options for clarity.
I did not just post poll answers on the original questions because users would be less likely to see all of the short poll options amidst the existing discussion and longer answers. Users who had participated in the thread could also not realize that there was any reason for them to return to the thread to express their opinion again in the form of a vote on a poll answer.
The use of the poll question was successful. Though the community's opinion was potentially unclear in the original question, in the poll question they indicated their preference for a particular tag name by a clear margin of 160 upvotes to 62 upvotes. This resulted in the tag name being changed.
In the discussion on the first question, one of the moderators declared that the tag name "swift" was not an acceptable option to him. The community did not express a clear opinion in that question, but the second poll question, they voted overwhelmingly in favour of the name "swift", and it was adopted.
From my perspective, the use of a second poll question seemed like a useful process wherein the community's opinion was clarified and manifested. However the moderator whose opinion was overruled did not agree, and he locked (not closed) the second question. When asked for an explanation, he said that "a poll like that was never a good, on-topic question". I disagreed.
In situations like this, is the use of a poll question acceptable? If not, what method is preferred for clarifying the opinion of the community when the existing state of the discussion leaves it ambiguous? As shown by this case, an ambiguous discussion does not necessarily mean that the community does not hold a strong opinion in a particular direction.