Should a dropdown always start with a blank option, or is a sensible default better?

Yes, required versus optional changes it, and in a way that catches people out.

For a required field, the blank option is doing real work: it lets validation distinguish "not answered" from "answered". Without it there is no unanswered state and your required check can never fire - which is the situation where the alphabetically-first country ends up in your database a thousand times.

For an optional field, blank is not a placeholder, it is a legitimate answer meaning "no preference" or "none of these", and it should be labelled as such rather than left empty. -- none -- or No preference is clearer than an empty row, which reads as a rendering bug.

The common mistake is using a blank first row as an instruction: an empty line where the label should be. If it says nothing, people do not know whether it is a choice or a placeholder.

There is a principle and it is sharper than case-by-case: does the field record a fact about the user, or a preference you are allowed to guess?

If it records a fact - country, date of birth, which account, how many people are travelling - a default is a lie you have written on their behalf. They will not notice, it will be wrong for a portion of them, and you will not be able to tell the guessed values from the chosen ones afterwards. Start blank.

If it records a preference where a wrong value is cheap and reversible: sort order, page size, theme: a sensible default is a kindness and blank is friction. Nobody wants to choose a page size.

The test that follows: what happens if this is wrong and nobody notices? Wrong country on a shipping form is a returned parcel. Wrong sort order is a click. That difference decides it, and it decides it consistently.

For question 3, the version that gets both: use the first row as a prompt, not as blank, and never as a value.

<select required>
  <option value="" disabled selected>Choose a country</option>
  ...
</select>

Selected so it shows, disabled so it cannot be submitted, empty value so validation catches it. The user sees an instruction rather than an empty box, nothing is guessed on their behalf, and the required check works.

And the genuinely both-benefits version for facts where you have a strong signal: pre-select, but make it visibly a guess. Detect the likely country, select it, and put a short line under the field saying you have guessed and they can change it. That gets you the low-friction path for the majority and an explicit invitation for everyone else, which a silent default never gives.

The part that matters: record whether it was confirmed. Then your data can tell the difference later.

Worth raising the thing above this question: how long is the list?

Once a dropdown has more than about twenty options, the blank-versus-default argument stops being the interesting one, because a long dropdown is the wrong control regardless of what is at the top of it. A search-as-you-type field beats it on every measure, and it removes the question entirely - there is no first row to argue about.

We have exactly this on a country field and the debate evaporated once it became a typeahead. Worth checking whether your disputed fields are actually in that category.