@JohnB
As I understood, this would be your first digital oscilloscope. I think it's a reasonable decision to go for a four channel machine. This would permit analyzing synchronous digital communications with separate transmit and receive lines (SPI for example). Since you mentioned microcontrollers / arduino stuff as one of your possible applications, you will likely need that. Alternatively, you could go for an instrument with two analog channels and an MSO option (logic analyzer) that provides protocol analysis on the digital lines. There's rarely the situation to really need more than two analog channels.
Since the price differences are quite steep, my recommendation would be to go for an entry level model first (this would set you back by somewhere between 300 and 600 USD). There's the Siglent SDS1000X-E series which is quite a modern platform with a good performance. The Rigol DS1000Z has been a class of its own for several years but nowadays may look a little dated, yet still a versatile and very affordable four channel scope. The GW-Instek GDS1000B series may be an option but at least in Europe it's a good bit more pricy than the others and hence I never looked closer at it, especially since I'm (too) well equipped with scopes... If you later find that you need something more capable, you will know better what features in particular you are looking for. You can sell the entry level scope with little loss or just keep it as a back-up unit.
To be honest, the instrument I still use the most, simply due to its very compact form factor, it's versatility and its lower weight, is my DS1054Z ("riglol'ed"
). In at least 95% of the situations it's performing as good or better (more modern firmware) than my other "main" scope, the MSO4000 (also "improved"). A few years later, I couldn't resist a really good opportunity to get a DS2072A-S from Rigol EU Clearance Sales (must have been 625 EUR IIRC) to replace an ancient TDS220 in my mess of a "basement lab". I was lucky this scope contained the full MSO hardware, so it was an easy "fix"... Compared to the old TEK it replaces, it's miles ahead but to the others, well, it just fits in the range. It's quite surprising that on the DS1000Z, Rigol adressed the (previously) piss-poor FFT, added signal filtering (math) options and made the derivative (math) function at least usable by firmware updates, while on the "upper range platforms" - well, meh..., just bug fixing (no, wait, the protocol decodes do work on memory and not only screen contents and also in zoom and segmented mode).
Business-wise, any of these scopes do serve me well (basically doing power, digital, analog and mixed signal control stuff, nothing high-speed, ultra-high-frequency or otherwise fancy), as probably would an SDS1000X-E or a GDS1000B or what not. The high bandwidth of the "liberated" MSO4000 is a fun thing (hobby, just like much of my other test gear...) but other than that, there's really little that this scope is doing better than the other models I've got. If I had to buy an oscilloscope for 3000+ EUR or USD today, I'ld probably have a close look at Rohde&Schwarz RTB2000 series, though I really hate that company for not offering the limited introductory version/pricing abroad...