By default the global set keyframe, the “k” key, is set to keyframe all Scoped channels.

However, if there are no keyframes on an object then hitting “k” at the object level will create keys on all the Translate, Rotate and Scale channels.

If I establish a key on the Y translate, for example, first; then move to another
frame and hit “k” then all the Rotate and Scale but just the Y Translate channels will receive keyframes.

If I change the Animation preferences for the Global Set Key to “Set Keys on Pending Channels” the only key'd channels that have values changed will received keys when hitting “k”.

Now hopefully you've followed this cause the question is this.
Should the Global Set Key automagically change to “Set Keys on Pending Channels” when an initial keyframe has been established?

I'll say yes, just from the point of view that fewer keys are better.

but I'm not an animator so I might not be aware of the best workflow.

It is set up the way it is now to allow for initial blocking then either the “Set Keys on Pending Channels” or preferably autokeying can be used to deal with specific parameter changes. I would recommend autokeying instead of changing how the k key works because then you can use autokeying for setting individual keys but still have the k key to block a frame if needed.

If you want to limit which channels are keyed on an object initially, you can change which parameters are Autoscoped. For instance if you only want to key translates, you can RMB on the scales and rotates in the parameter pane and select Channels and Keyframes > Unset Autoscope. Now when you select that object then only the translates will be scoped and keyed.

Let me expand on Rob's comment. “Set Keys on Pending Channels” is problematic when blocking. Let's take the simple case of animating translation channels with this option turned on. Suppose for the post at frame 1, the animator created channels and keyed TX, TY, and TZ. Then at frame 12, only the TX, TZ channels were keyed by hitting “k” because only those channel values changed. Then at frame 24, the TY channel was keyed because only that channel's value changed. Now due interpolation and the fact that a key for TY was not set at frame 12, the pose that the animator hit “k” at frame 12 is now different from what the animator had hit “k” on. The animator might not even realize this until later when they do a flipbook, and by that time, they probably had no idea why the software had “!@#!% screwed” up their keyframing.

This example was only on translation channels. The effect is much more dramatic if the example used rotation channels because for orientations, the RX,RY,RZ channels have interdependencies.