- A safe streaming library where effects are guaranteed to only be performed once: https://www.tweag.io/blog/2018-06-21-linear-streams/
- Efficient zero-copy network communication: https://www.tweag.io/blog/2017-08-24-linear-types-packed-dat...
- Safe manual memory management and prompt resource release: https://www.tweag.io/blog/2020-02-06-safe-inline-java/
- Explicit API design to guide developers on what function to call next when implementing a protocol like TCP: https://www.tweag.io/blog/2017-08-03-linear-typestates/
- Add correctness proofs to already existing algorithm implementations like merge sort: https://www.tweag.io/blog/2018-03-08-linear-sort/
But we already had ST for that, so the real advantage of LTs over ST is you can type-safely "freeze" and "thaw" your mutable structure without copying.
The mutable array example in the Linear Haskell paper is a simple comparison to what you're asking for.
Its hard to find real world use examples and learn about linear types when I search online. If anyone has links, that would help.