When there is not a prescribed use case for a picture, it makes sense to compose the result such that symmetric cropping to other image ratios (which most automatisms forcing material to a certain format will perform) will not result in a loss of core image information. For example, the following 4:3 image does not work well for a symmetric crop in 16:9: so it would have been better to compose it in 16:9 (note that if you are photographing in JPEG+RAW mode, most cameras will retain the full image information in RAW so that composing for a crop does not throw information away that other crops might still need).

So this image wants salvaging by manual cropping. However, in this particular case making the 16:9 crop the "canonical" crop is not a good idea since it is vertically overfull. Too much spider. It's an emergency composition when you need 16:9 (because the original framing left too few reserves). If other crops start from the 16:9 crop, they will be equally crowded vertically.

So what you do instead to create a "canonical crop" for this image is just cropping enough off the bottom of the 4:3 image that a centered crop to 16:9 will work out as well as it can. Then it's only the 16:9 crop that will be crowded, with other crops retaining larger top and bottom borders.

Without the specific problem of crowding, the "canonical crop" might well tend to be 16:9 for images with mainly vertical content (then all further crops take from the sides) or 4:3 for images with mainly horizontal content (then almost all further crops take from top/bottom), assuming landscape. For portrait mode, considerations are similar.

Resizing is a step left for finalizing the material to its target orientation, resolution, and framing. If you are not asked for it, don't do it. If you know the ultimate aspect ratio to be used, prepare a crop with the right framing and the right aspect ratio. Then the finalization will just involve scaling to the final resolution (often left to be done by the print device).

For material where you don't know which aspect ratio it will be used with (like screen backgrounds), the approach is to assume that the final aspect ratio will be achieved by symmetric cropping from the axis containing too much material: that's the straightforward dumb strategy that you can prepare for: anything smarter should be able to work from there well enough. The rest of this answer caters to this scenario.

When there is not a prescribed use case for a picture, it makes sense to compose the result such that symmetric cropping to other image ratios (which most automatisms forcing material to a certain format will perform) will not result in a loss of core image information. For example, the following 4:3 image does not work well for a symmetric crop in 16:9: so it would have been better to compose it in 16:9 (note that if you are photographing in JPEG+RAW mode, most cameras will retain the full image information in RAW so that composing for a crop does not throw information away that other crops might still need).

So this image wants salvaging by manual cropping. However, in this particular case making the 16:9 crop the "canonical" crop is not a good idea since it is vertically overfull. Too much spider. It's an emergency composition when you need 16:9 (because the original framing left too few reserves). If other crops start from the 16:9 crop, they will be equally crowded vertically.

So what you do instead to create a "canonical crop" for this image is just cropping enough off the bottom of the 4:3 image that a centered crop to 16:9 will work out as well as it can. Then it's only the 16:9 crop that will be crowded, with other crops retaining larger top and bottom borders.

Without the specific problem of crowding, the "canonical crop" might well tend to be 16:9 for images with mainly vertical content (then all further crops take from the sides) or 4:3 for images with mainly horizontal content (then almost all further crops take from top/bottom), assuming landscape. For portrait mode, considerations are similar.