Zero-width space (U+200B) characters found embedded in release-8.5 English content

[yahonda]

What did you do?

While working on PR #23778 (removing stray zero-width-space MT artifacts from the i18n-ja-release-8.5 Japanese translation branch), I checked whether the same artifact exists in the English release-8.5 branch and found that it does.

What did you expect to see?

No invisible/zero-width characters in the English source content.

What did you see instead?

45 occurrences of U+200B (zero-width space) embedded in prose across 16 files in release-8.5, most commonly as a doubled zero-width space between two words (visually invisible, but present in the raw text), e.g.:

• literal-values.md: "the literal values​​of TiDB SQL statements" / "For month or day values​​less than 10"
• statistics.md: "the number of values​​falling in the bucket" / "the number of values​​falling into each bucket"
• dashboard/dashboard-diagnostics-report.md: "configuration values​​of historical time" / "the sum of all values​​(SUM)" / "P80 values for a single transaction of KV writes​​are"
• Also present in: develop/dev-guide-join-tables.md, encryption-at-rest.md, constraints.md, information-schema/information-schema-deadlocks.md, information-schema/information-schema-data-lock-waits.md, information-schema/information-schema-tidb-trx.md, dm/dm-binlog-event-filter.md, dm/dm-block-allow-table-lists.md, and several releases/release-*.md files.

These are invisible in rendered output, but they exist in the raw Markdown source and get carried into any downstream text processing (search indexing, translation tooling, diffing, etc.).

Question

Are these zero-width spaces intentional (e.g., inserted by some tool for a specific rendering purpose), or are they unintentional artifacts (e.g., leftover from a word processor, browser copy-paste, or an editing tool) that should be cleaned up? If unintentional, I'm happy to submit a PR removing them.

Which version(s) do you use?

release-8.5 branch (also worth checking whether other release branches / master are affected — not checked yet)

[yahonda]

Opened #23793 to remove the zero-width-space characters from the 13 files where docs-cn does not have the same characters. 3 files (dm/dm-binlog-event-filter.md, dm/dm-block-allow-table-lists.md, releases/release-6.0.0-dmr.md) were deliberately excluded from that PR since docs-cn has the identical characters there too, which might indicate a shared/intentional origin worth investigating separately before touching those.

[yahonda]

Traced the origin: all 45 zero-width-space occurrences across all 16 affected files were introduced in a single commit, aa389a3 (PR #23696, "docs: use ticdc-version variable in TiCDC examples"), confirmed via git log -S"​" -- <file> for each file.