When I cleaned records using a Loricraft point-nozzle vacuum, the most critical step was the rinse. For that I used Nerl Reagent Grade Water from Fisher Scientific - it came in 5 gallon (44lb) boxes and was expensive, largely thanks to shipping costs. Reagent Grade water is primarily used in biological laboraties and others where ultra pure water is required for test integrity, etc. It works great for cleaning records but I've come to the view that it overkill and too costly. If you have a university near you, go to their Biology Department who probably makes their own; they may be able to supply you with some.
For ultrasonic cleaning, distilled water is fine. Get a TDS (total disolved solids) meter, $10-$20, and test your water. Distilled water should test at 0000ppm (parts per million), indicating no dissolved solids. Adding IPA and/or a wetting agent should still yield close to 0000-0001ppm.
That's great for a fresh round of record cleaning. BUT, the water will get dirty from cleaning records. If you air dry (fan or natural evaporation) it is the dirt in the water on the record that will remain on the record. Use that TDS meter at the end of a cleaning cycle to see where the water is now. There are 2 ways to deal with this: i) multiple rinse steps with clean distilled water or ii) actively filter the water in your ultrasonic cleaner.
It's all covered in this article:
tima’s DIY RCM – follow-up #2: Compelling Changes – Improved Results
While it is important to start with clean water, it is the water on the record when it dries that matters most.
Here is an article on
Differences Between RO and DI Water for Parts Cleaning, Washing, and Rinsing
[DI water is deionized water]