First off, the logic that let you do this is still in place - in fact, I just went to Cross Validated and cycled through several items in the queue using exactly the technique you described.
...but it won't always work. Here's the logic that review uses to give you a "random" review task when you visit /review/close:
- Construct a query that'll return all reviewable tasks for you, excluding your own posts, those you've already reviewed, etc.
- If you're filtering by specific close reason(s) or tags, then restrict the results of #1 such that only those in specific close reasons or tags appear.
- Take the first 1000 matching review tasks (newest-first) OR the first 10 matching review tasks (if you're filtering by close reason) OR only the very first matching task (if you're filtering by tag).
- Attempt to filter the results of #3 by your tag preferences (or the tags you've been recently active in).
- Pick a random task out of the results of step #4 (if filtering was successful) or #3 (if it wasn't).
Now we can make some educated guesses as to why it wasn't working for you earlier today:
- If you were explicitly filtering by tags, you'll always get the newest task that matches one of those tags.
- Even you weren't explicitly filtering, if exactly one task in the queue matched your tag preferences on the site (in favorite tags and not in ignored tags), you'd always get that one back.
- Even if you have no explicit tag preferences, it'll fall back on the tags you're most active in and try to give you a task in one of those.
- Even if none of the above applies, and you do actually get a random selection from a large pool of tasks... That selection is based directly on a random number generator, not a fair shuffle - so it's remotely possible that it will return the same number several times in a row and thus give you the same task.
My money would be on the middle two options there; if (for example) the system thought you really like [regression] questions and there happens to be exactly one [regression] question in the queue... You're gonna get that question every time until you skip or review it.
I should've mentioned this earlier, but... If the only reason you're not using "skip" is so that you can come back to it later, then there's another way of doing that, provided you have at least 10K on the site: you can just go back through skipped items in your own review history for that queue. For a small queue, this is probably not much more work than relying on random selection.