UriUtils.encode("a&b=c", false) returns a&b=c, whereas the corresponding call with an existing triplet, UriUtils.encode("a&b=c%20", false), encodes the raw reserved characters. The no-triplet branch in encodeInternal passes a hard-coded true to encodeChunk instead of the argument.

Expected: a%26b%3Dc when the flag is false; existing %XX triplets should still be preserved. The true-flag and flagless overloads keep their current behavior. This concerns the existing boolean contract, separate from #2002's discussion of double encoding.

Reproduced on current master with UriUtilsTest: 32 tests, four failures, no errors or skips. The failing cases include both boolean overloads and non-ASCII input. Main sources compile with JDK 8 and tests run with actual JDK 25 using the project's dev profile.

I will pass the flag through in that branch and add these regressions. The scope is UriUtils, its existing test suite, and the changelog; no URI-template expansion or query collection changes.

_{Reproduction and report prepared and posted by an AI assistant for @hutiefang76.}