Query Loop: editor unresponsive + IP blocked after any block setting change

We’re experiencing a reproducible issue with the Query Loop block on GenerateBlocks 2.2.1 / GenerateBlocks Pro 2.5.0 (GeneratePress 3.6.1, WordPress 7.0, PHP 8.4.23, hosted on Raidboxes). Whenever we edit almost any setting on an existing Query Loop block in the editor — including simple, non-taxonomy changes like increasing “Posts per page” from 10 to 14, or opening the “Select taxonomy” dropdown — the editor becomes unresponsive shortly after (“Updating failed. Could not get a valid response from the server.”), and our IP is subsequently blocked from reaching the entire site (ERR_CONNECTION_REFUSED) for 5–10 minutes.

This happens consistently, including on a freshly cloned copy of the site, which might rule out any content- or data-specific cause on our end. Our working theory is that the Query Loop block fires REST API requests on every settings change without debouncing, and that this request pattern is triggering our host’s rate-limiting/firewall layer. Could you confirm whether settings changes in the Query Loop block are expected to trigger a REST request per keystroke/change, and whether there’s a known issue or setting to reduce this request frequency? Happy to provide server logs, or screen recordings if useful.

For additional technical context: our server logs show the Query Loop editor firing repeated POST /wp-json/generateblocks/v1/dynamic-tag-replacements requests in rapid succession — e.g. 17 separate requests within the same second (16:57:51–16:57:52) while the block was open in the editor, all returning 200 but with no apparent debouncing between them. Separately, we’ve also observed unresolved dynamic tags leaking into the rendered frontend HTML for this same post type — e.g. a request to GET /{{post_meta key:article_pic}} (404), suggesting the tag {{post_meta key:article_pic}} was output literally instead of being replaced with the actual field value. Not sure if these two symptoms share a root cause, but flagging both in case they’re useful data points.