Are there any fonts that try to compose Unicode Ideographic Description Characters into one character? E.g., ⿰女子 would look like 好. I know the Wikipedia page says "rendering systems are not intended to automatically compose the pieces into a complete ideograph", but this could be helpful for scripts like Sawndip that aren't fully encoded in Unicode. Composing characters overlaid with ⿻ would probably be impossible, but it seems that it shouldn't be hard to at least try to compose characters that are put together just side-by-side with ⿰ ⿱ ⿲ ⿳, especially since something like this is already done for Old Hangul syllable blocks.
Let’s turn this into an answer.
If you wish to create arbitrarily new characters, give GluphWiki.org a try. Make a version in their editor (not flawless, as you’ll have to physically stretch and resize components, but better than lots), save and then extract as one-glyph font. (The symbol will be at GETA MARK [U+3013].) Only watch out for their naming practices and be careful to keep to the national preferences. These glyphs look best with Hanazono fonts, which are basically pre-made from GlyphWiki.org, so try it on. Especially the Hanazono-AFDKO (Google it) versions, as they are well-versed in setting up the variants.
If, however, your wish is to recreate some characters really existing but not in Unicode, first scan BabelStone Han PUA font, it might be there. Works best with BabelStone Han (actually, is part of it), so you might want to install that. Note: only includes PRC forms.
The actual composer of IDS (for lost of well-known non-Unicodes) is the Hanazono Mincho I from the aforementioned AFDKO. Scan its proof (https://github.com/cjkvi/HanaMinAFDKO/releases/download/8.030/HanaMinI.proof.pdf) to learn which are supported. Note that the most well-popularized examples of non-Unicodes are sometimes even in mass-use fonts, such as ⿰氵恩 (only unencoded Korean Name) or ⿳雲⿲雲龍雲⿰龍龍 (taito) are in Source Han sans.