A single flow cell contains a few thousand pores (I think this is what you mean by "holes") that are all at different stages of passing different molecules, with signal data being captured from a few hundred at any given time. In practice you'd never expect (nor could you arrange) for two pores to be at the same stage of processing the same (or any pre-determined) molecule at the same time, so correlation information like that is out. The "clock rate" is determined by the so-called motor protein that "pulls" the nucleic acid molecule through the pore, if you fancy going down the reading rabbit-hole...
I used to run a department at a biotech where ~50% of our data came from MinIONs (although, that said, I'm a bioinformatician, rather than a molecular biologist), so I can answer your questions. For (a.), you can for sure "batch" samples. The term of art you're looking for is "multiplexing". Nanopore provide prep kits that allow you to "barcode" different samples (i.e. tag all the molecules in a given sample with a unique, synthetic sequence, which allows them to be distinguished by software downstream), but note that (as with all DNA prep kits, but some more than others) you'll need access to a fair whack of lab equipment and consumables to use it (these kits aren't "all-in"). For (b.), for one anecdata point, I used to process a whole flow cell's data on an M4800 with a 4th Gen i7 and 32 GB of RAM in a few hours. Most of the "high" computational requirements you hear about relate to either assembly or variant calling (both of which are downstream of just retrieving "usable" sequencing data); and even both of those I've managed on that same laptop overnight. Actually acquiring the data (you can delay base calling if you like, although you probably wouldn't need to) is real-time and only needs very modest hardware (IMHO the Nanopore "system requirements" are very much on the "safe-side".) "In the field", your challenge would be physically preparing the samples!
This is a common, and often justified, though not always fair, criticism. MinIONs have an error rate of around 10% for _any given base_. Moreover, these errors aren't entirely independent of one another, so if you struggle to sequence a given base the first time, you're likely also to struggle if you try again. That said, if your experiment is such that you're only sequencing a guaranteed single target (e.g. one, isolated coronavirus genome), in that one sequencing run (on that one flow cell), you'll "re-sequence" the same any given region many times and, unless you're looking at "problematic" (i.e. low-complexity) regions, you _will_ be able to "average out" the errors to reveal the true target sequence. On the other hand, if you're trying to co-sequence a mixture of closely-related targets, that's when the headache starts...
It's a complicated issue; I tend to think of the error component of any one MinION observation as being a function of the k-mer in the pore at the time (i.e. the subject of the observation) and, with some decaying dependence, the sequences (i.e. in both directions) that extend out from either side of the target k-mer. You might say that MinION error is a function of the target k-mer and its immediate environment. It gets even messier when you try to imagine the form of that function; for one, it's not _completely_ good enough to remain in sequence space alone: among other things, the "shape" (i.e. the conformation) of that (DNA or RNA) molecule around the target k-mer will influence how the shape of the pore will change in response to the target k-mer, which, in turn, will influence the observed current signal (i.e. manifest as a deviation from the "expected" or "ideal" current signal for that k-mer!). As I understand it, Nanopore don't spend too much time actually modelling k-mer-in-pore dwell-mechanics; instead their best base callers use machine learning to generalise across the swathes of available sequencing data for known targets (and give really quite impressive results, all things considered).
As others have said, you're reading a sliding window of k-mers over the target sequence; I think for the MinION k is presently 5. To answer your question directly, it struggles with homopolymer runs, not inherently because they're low complexity, but actually because it's tricky to "clock" how many like, contiguous k-mers have passed through the pore after a given period of time. That is to say, for example, if your target sequence is "GGGGGGG" (i.e. a homopolymer run of 7 Gs), you'd expect to observe three like, contiguous signals (i.e. in current space) for the all-G 5-mer, one signal each per "clock cycle" (which corresponds to the dwell time of the k-mer in the pore). If these "clock cycles" were always constant, it's merely a case of dividing the "time spent on the observed all-G 5-mer" signal by the the "time spent on one clock cycle". Sadly, for our purposes, there's enough wobble in any one such "clock cycle" that that calculation won't always yield a reliable result. The upshot: your "GGGGGGG" (7 Gs) target sequence may be registered as "GGGGGG" (6 Gs) or "GGGGGGGG" (8 Gs), or even something else. Now, for distinguishing two alleles where the difference between them is, say, a doubling in length of an already-very-long homopolymer run, even with the aforementioned "clock wobble", you'd likely be able to see that in MinION data quite clearly. As with all thing DNA sequencing (for the time being, at least!), your precise biological question will determine which (one or more) sequencing techniques are best for the job!