Canvas boards with very large cards are refused up front, and long links render faster
A canvas whose cards would take seconds to lay out now gets an immediate 400 `canvas-too-complex` saying what to change — split one huge note, or use fewer cards — instead of holding the request. Shrink-to-fit is much cheaper and long words break with a search, so ordinary boards render in the same time or less.
POST /api/v1/canvas now estimates each card's layout work from its
converted content and refuses a board over the budget before that card
is laid out. The response is a 400 with code: "canvas-too-complex",
rule: "layout-work", and details naming the card that took the board
over — the same self-correcting shape as the existing
markdown-too-complex answer.
What counts is layout work, not text length: thousands of inline elements (code spans, links, list items, headings), one whitespace-free run far wider than a line, or a card so short that its text spans hundreds of one-line pages. The budget is board-wide and charges a small fixed cost for every card, so a board of many ordinary cards can reach it too: 1,000 one-line cards still render, but a few thousand tiny cards is a very large board and gets the same 400. The response depends on which side is over. One huge note gets "split its text" — long notes belong on their own canvas or across several smaller cards. A many-card board gets "use fewer cards": merge small cards into larger ones, or split the board across several canvases.
Two layout passes also got cheaper for everything that stays under the budget. An overflowing card now probes the first shrink step and then the 60% floor, so a card whose content can never fit costs three layouts instead of nine; a card that can shrink still shrinks to the same size as before. And a long unbreakable word (a URL pasted into a link node, say) is broken with a search rather than a measurement per character, which makes a 40KB link card several times faster to lay out.
Details are in the canvas limits section of the API reference.