I think this is best explained by an example. I used to work for an analytics company working on a tool that was designed to identify, unmask, and track website visitors. The end goal was for the website/app owner to know exactly who is visiting their website/app, including information such as their full name, social media accounts, email address, phone number, etc. The way some of these tools are built is to fingerprint a user, including their device, browser, and more importantly data which they submit. Once the name is submitted, you can set cookies, session storage, or local storage to continue to identify the user across different websites. If the tool is integrated into 2 websites, it could capture the information about the user on one site and then provide that information to the owner of the other website.
The more information you have on the user, the more accurately you can identify him. If the user wipes their cookies and cache, you could still track him if his fingerprint is incredibly unique. For example if it is unique out of the 1000 users, it’s highly probably that it’s the same the user when you encounter the exact same fingerprint. The more data you collect on the user, the better you can identify him.
Whether you go incognito, use another browser profile, or a new tab doesn’t really matter these days since tracking is no longer done just by cookies alone. For example, if you visit this site https://fingerprint.com using Arkenfox and then Librewolf without any change in IP address, fingerprint.com will identify you as the same person. You can try doing the exact same by opening another tab.
One reason I recommended using SOCKS5 proxies in another topic (Browser Use Cases - #3 by vergeOfNormy) is because a change in your IP address can throw off the analytical tools and you will no longer be identified as the same user.
Feel free to test it yourself.