Remains the improvement it would do on the AI as the motivation to remove HQs.
So...
Army HQs will remains. Because they are few of them, so keeping them don't add a significant amount of micromanaging. Also, they are not to be created on the fly, just because a 3 stars general desire so.
Divisions HQs will be removed. Because they are numerous, and this add micromanaging. You will need a general, you will have a maximum limit of divisions at a given time, that can be increased by events. The leader will pay the formation of a division by having reduced stats during one turn. This will cost you war supplies. If the leader is removed for any reason, the division is dismissed and the brigades reappears.
The side benefit will be also to be more consistent with the Chain of Command, from a design standpoint. Many people wondered why there was a 'hole' in HQs (no corps HQ, but army and division ones), naturally, with the Div HQ gone, there is no hole to consider

This is also rather historical in our mind, to not have Div HQ. Divisions at these times were organized far more rapidly compared to forming a real Army.
We will also add (and this will be the first thing done) a new special order: 'Relocate'. This order will be restricted to leaders and Army HQs, and will allow them to redeploy instantly within a reasonable range. The number of usage of this order will be set at 3 usages a turn per faction. This order will provide several benefits:
a) the AI will use it to relocate instantly Army HQs where one is needed, thus solving the last part of the problem about synchronizing an HQ with a leader.
b) players complaining how non realistic it is to have a 15 days travel, by train, for a single general, between say Charleston and Richmond will be satisfied here. (and the complaint is valid).
Thanks all for your participation. We can still discuss here the consequences of this decision and the thread is not closed.
To people thinking that it is abnormal to redesign an important part of the game after release... I would like to say that we take, Philippe and I, many decisions during the development of our games very instinctively, and 95% of the time we fall right where we wanted. But we also have to admit that we are not error proof. In this case, we don't feel we are discussing an 'issue' per se. HQs works fine as they are. But we wanted to simplify a bit the rules, because some peoples had trouble with it, and we also wanted to boost the AI expertise. With time and much work, the AI could have understood perfectly well how to position HQ, but we are feeling more important improvements could have been done with the time spared.