From a purely theoretical perspective, when you are a baker who is capable to reaching 100% target of its priority 0 baking rights, the optimal situation in terms of your capital is to have all of it blocked as security deposit because it means that all your capital is working. In such situation as soon as some bond is unlocked it would get relocked again by a new baking and at the same time you would never miss a single block/endorsement baking right. To be in this situation means that the baker is exactly 100% delegated at all times: there's just enough delegators pointing to you as your capital can absorb. It is the optimal situation for both the baker and the delegators.
In reality, bakers are rarely in this situation, in practice they are either "under"-delegated or "over"-delegated.
If you are under-delegated: it is suboptimal for the "baker" this means that you could have accepted more delegations and received more baking rights without impacting your success rate, your personal ROI is suboptimal. On the other hand this situation is fine for delegators because they still receive 100% of their expected rewards on average for their level of staking (minus the baker's fee of course), their personal gross ROI is optimal (assuming you don't make any operational mistake while baking of course).
If you are over-delegated: is is suboptimal for the delegators because your capital is not enough to cover for all the baking rights of the operation. There's less rewards coming in the bakers' account than the expected amount to pay all the delegators. In such situations the baker usually pays himself first (his personal ROI is optimal) so he received his expected value however the effective amount received by delegators will depend on the policy of the baker (pay those who came first in priority? Or distribute the loss equally to everyone?). So in this scenario the baker rewards are usually "covered" by the baker policy but not the delegator's rewards (the delegators gross ROI is suboptimal).
So all in all, given that delegators often enter into a delegation and do not monitor day to day situation of the baker's bond and pay a fee to that baker, they usually expect the baker to do this monitoring job and give them a heads-up if they believe they might get over-delegated in the near future in order to give appropriate time for delegators to find an alternative.
For example in the beginning of mainnet, as per the protocol the initial bond requirement was kept intentionally low (started from 0%) and gradually ratcheted up over the cycles. This means that many bakers have been to start their operation with quite small actual bond requirements which allowed them to ride the opportunity and accept much more delegations than their capital would allow once the bond requirement would achieve its long term level (512XTZ per block and 64XTZ per endorsement). This created a situation where a number of baker got not immediately over-delegated but forward-over-delegated.