Workflow depends on lots of things...
When you resize a frame in InDesign, the content is unchanged. This may seem counter-intuitive, but the only thing that changes is an internal descriptor which controls the dimensions of said content. For example, if you have a raster 5 megapixel image, it is 5 megapixels regardless of frame size. Thus, there is an "appropriate" size that depends on your final output. If you are printing a glossy art book using a 200 lpi halftone, your publisher or printer may want a minimum of 300 to 400 ppi effective resolution. (This is based on the industry rule-of-thumb of 1.5 to 2 times the lpi; see this answer.) Resizing a frame containing vector content is the same, it merely changes a notation that the content should be reproduced at a given dimension. In both cases, the preview that InDesign generates may be of varying quality (based on preferences), but of course shows the content filling the frame accordingly.
The main effect of placing large content and resizing it in InDesign is processing times downstream. I'll use an example I dealt with all the time: Imagine an 8x10 inch image was scanned at 600 ppi, resulting in a ~29 megapixel image. If it contains 32-bit pixels (CMYK) it would be about 920 MB uncompressed. It's placed in a small 2x2.5 inch frame. InDesign still sends the 29 megapixels to the printer, PDF-creating software, platesetter, RIP or whatever. If the job were to print a basic grayscale picture at a low lpi, say 100, then the time needed to transmit and process nearly a gigabyte of data would be a significant waste compared to the minimum necessary to produce the desired quality. 2x2.5 inches at 150 ppi requires 112,500 pixels, which in 8-bit grayscale is only 900 KB. That's several orders of magnitude of difference in data size!
The consequence of sending 1000x more data may or may not matter for one image, but when you're publishing a magazine or book, dozens or hundreds of images with such excess data can really add up. It also depends on hardware. When I started in printing, it took eight hours to send twelve CMYK images to an imagesetter for a calendar. Today those same images likely could be chewed through in mere seconds or minutes.
I should note that InDesign can send only the data that is needed, based on settings for your output device, but it was my experience that this sometimes led to unexpected results.
Should you go through and size every image perfectly? Probably not. It depends. If the image is too large but only by a little, it's not worth your time (or your employer's time) to resize it. If a publication is absolutely massive and going to cause problems, we would put the burden of resizing images back on the customer; or impose a fee for additional processing time. If a job is going to take four hours to send to the platesetter (and should only take minutes), versus two hours for me to resize images properly, I resize them. My time was cheaper than the platesetter. Can you (or your publisher) afford the time on an expensive machine that isn't needed?
On the other hand, if you are designing for a client that comes back to you and says "We changed our mind, delete the copy on page 23, swap it for page 24 (the color folio), make the photo fill the page!" When that happens, you better hope you didn't overwrite the original massive file with your tiny grayscale one! As you can see, it depends on your workflow.
As a designer/prepress technician, if I had to resize images to make processing smooth downstream, I would create a backup copy of all the client's original images, resize the working copies, and send that. Storage is cheap, so having massive files on hand for future changes is useful, but tying up a digital press, platesetter, or whatever isn't. Once the job is approved, printed, shipped, and done, you might have the option to delete the backup copies.
InDesign sends the full data for the link. Downsampling occurs at the output device. See #2.
Vector output resolution is determined mathematically by the output device. InDesign just sends the original content. See #2.
If you know exactly what size and resolution you need, ideally content is already sized appropriately before placing in InDesign. However that's not always possible. That being the case, it should be resampled before output to printer/RIP/platesetter, etc. Whether resampling is worth your time or not depends on how much time it consumes on the output device, or how much needless data transfer and storage it causes.
This is where backup copies come in useful! (See #2.) If you resample something and later discover a different size is needed, you need to be able to go back to the original. In almost all cases, original files should be at the highest resolution that they might be used for. If your client is submitting an image that is being used on a color flier as well as a business card, it should be at least the size and quality needed to make the flier (the larger output) look good. You can always downsample for the business card, but you can't "upsample." (Well, you can, but only to a point, and only to help avoid making images look overly pixelated.)
Adobe has a workflow solution that allows you to store revisions of documents such as InDesign and Photoshop files. If your company does a lot of design and creative work, you may want to look into using it. It allows you to revisit past versions of all documents much the same way a source code repository works. It helps you if a client can't make up their mind and changes images or other elements of a document often. (I haven't personally used it, so I can't vouch for its utility.)
I'll assume by "liquid/alternative" you mean different versions of the same publication, wherein images may be different sizes, copy may be different languages, etc.
The most experience I have with this sort of thing is where the client has images and copy for a print- and web-based version of something. In such cases, sometimes they submit images already sized for both uses, but often they simply provide the original images from the photographer or studio. It was my job to resize them appropriately for each publication.
In the case of web versus print, our firm would create a shortcut to the original photos in separate working folders on the server. The web designer would use them as source material as would the print designer. Depending on circumstances, we might create a "working" copy of the original image in the original folder. For example, a photo of a real estate agent, provided by the customer, resides in an "Original" folder, and everyone knows not to alter them. The customer requests some retouching and maybe some color correction. The new version of that image (still at the original resolution) is saved to a "Working" folder inside "Original." The print and web designers both use those versions of the images, downsampling as necessary for their local version of the publication. If any changes are made to the working copy, both designers would have to update their local copy again.
A better workflow is to have all designers use the "working" copies (even though they are probably too large), and wait until everything is approved before creating downsampled local versions.
In a separate example, say you are making a brochure in two languages. You can use layers in InDesign to have both versions in the same document, of course. If the images are sized differently, I typically create a resampled version of each. If the original image was "Doctor Smith CMYK.tif" then I would probably create new files named "Doctor Smith CMYK (English).tif" and "Doctor Smith CMYK (Spanish).tif". It can be troublesome to remember to recreate these new files and re-link them if the original is modified by the customer.