For manual testing it would be smartest to re-use the page (or test) state, because it will be way faster. It is possible to argue the same for automated tests, but if you do not run tests in isolation from each other there is a high risk that you will loose confidence in your test suite. This because failing tests have possible side-effects on each other that cost a lot of time to investigate if you can reproduce it at all. Certainly when one test fails there is a high chance multiple will fail as they are depended on the state of the page. Which test will you research first? How do you know you have the correct one?
Automated end-to-end tests are there to test the product integration and that common workflows still work. It is not a replacement for duration tests or how feature work together over time. Create different tests for this.
I would advice to not re-use pages or state if possible. I have had a situation where workflows where very time consuming. In the end I created a lot of small tests which at the end of the test would save their state to a database in order for it to be quickly re-used by depended tests. This way we could run each test in isolation, but still re-use the application state if the test was successful, speeding up the whole regressions suite greatly. We documented all the tests dependencies in code so it was pretty clear how to research fails.
Maybe now test execution time is not an issue, but it will be as your product grows and market cadence expects faster releases. Being able to run your tests in parallel is a must. To reduce test execution time now and in the future it is better to start with writing tests that can be run in parallel from the start. If you do not do that it is so much harder to scale in the future. State-reuse therefor is an anti-pattern, because it will limit running tests in parallel in a later stage.
In the end it depends. Preferable you run tests in parallel instead of re-using state to make them faster. Sometimes this is hard, but do try to find a way that tests do not influence each other to minimise research into failing tests.