Variable damage is an interesting idea. Just on the topic of variable damage, there are many readability and discoverability issues to solve.
How is the variable damage reflected in the explosion effect?
How is the damage that a Skuttle will do reflected on the unit?
How do players learn about this mechanic?
Basically, you need to imagine that you are learning Zero-K purely through observation. How closely do you have to observe the game to learn that variable damage exists? How long until you figure out the mechanics? Take Scorcher as an example:
The variable damage is very stark. Scorcher deal tickle damage at range and massive damage close in.
Multiple data points can be gathered locally in time and space (since you can watch Scorchers damage change within a single battle).
The projectile light and thickness falls off with range.
Scorchers are one of the most common units.
All mechanics have discoverability costs, but I think Scorcher has many more ways to ameliorate this cost than Skuttle.
Anyway, the readability and discoverability issues inherent with variable damage are solvable. Note that the solutions would take quite a bit of work. There is definitely a design cost to adding variable damage, but relative to the current Skuttle design it may be worth it.
Health Variable Damage
Now, onto your specific implementation of variable damage. I don't like basing damage on health because it seems like a unit AI and UI nightmare. As written, your proposal would require players to press the detonate button to deal any significant damage. This requirement will not be clear to players as it runs counter to how every other bomb works. They will simply let their Skuttle die, see that it does very little damage, and learn that Skuttle sucks.
Players that know the theory behind Skuttle are still left fighting against the UI. It is hard to judge the kill radius of a Skuttle against various units if the Skuttle is losing health. The actions required would be very quick and lag-dependent. If a Skuttle is about to be hit for a large amount of damage then it may as well detonate just prior to taking damage.
A decent Skuttle AI would be quite hard to specify, and I try to avoid adding mechanics only to effectively remove most of their scope for player interaction with AI. For example, I can't add the exact Spy mechanics from Red Alert 2. For a positive example, the skirmish AI can exist because a large part of the player interaction (whether to dive in or retreat) is preserved. I fear that a Skuttle detonation situation would play out too quickly for any manual overriding of a Skuttle AI to be feasible.
Idea #1 seems too similar to Jack, depending on how much you want to buff its speed. If Skuttle cannot cloak then it can only assault light to medium defenses, and does so by tanking damage until it can jump into melee range. This is exactly how Jack works against defenses.
Mobile armies can run away from a non-cloaking Skuttle unless it is very fast or can jump quite far. Limpet is almost 3x faster than the current Skuttle and can often reach enemy armies before losing 400 health.
The burrowing in Idea #2 is a bit more interesting. A jumping burrower feels like a unit with more deliberate use. Some UI that replaces attack commands with predictive jump+detonate commands would mitigate many of the UI issues.
Alternate Variable Damage Proposal
Add the following variable damage rule:
Skuttle can charge up its damage over a few seconds, to some maximum value.
Charging or maintaining charge decloaks the Skuttle.
Skuttle automatically charges whenever it is decloaked.
Skuttle detonates at its current charge level when killed.
The rest of the Skuttle redesign can be buffs to offset this nerf.
This has similar features to your proposal, but fixes some issues. They are most similar in that the defender can mitigate some damage by dealing a large amount of damage in a short period of time. In your proposal this damage 'counts' if it is dealt en-route to the target, whereas in mine all damage that shortens the Skuttles life counts.
My take on variable damage fixes many of the discoverability, UI, and AI problems.
Skuttles can visually and (possibly audibly) sparkle and look fancy as they charge up.
New players that are unaware of the Detonate button are taking close to optimal actions against armed static target.
The only reason to explode early is if the defender runs away from the Skuttle.
Force Fire and Detonate can be changed to tell Skuttle to charge up and explode at maximum charge. Self-D can be the command to instantly explode.