Sometimes Caches get corrupted with bad data, e.g. if the cache is populated with data from an external API which has returned bad data, it might be necessary to flush the cache.

In such scenarios it may not be feasible to wait for the cache to time out.

In Azure Preview Cache both co-located and shared (on own instances) the max-item size is 8 mb per item.
Since its a dedicated cache the max limit of 8 mb should be configurable (for example for session state usage).

Federations are an excellent feature, but their scale-out potential is hindered by limited functionality. Without indexed views, the switch to federations for many applications would require significantly more resources to see performance gains, if any.

Add a feature which would allow me to set a cap on pay as you go plans. I should be able to set a trigger point for expenditure which will notify me (email, portal notification or both) when it is hit. I don't think my account should be shut off or suspended, but it would be good to keep tabs on how much we're spending.

When a queue suddenly goes from being empty to having an item, or from having items to having none, it would be great to be able to spin up, or spin down a role.

Some protocols (most notably FTP in PASV mode) require creation of temporary single-use inbound endpoints. Allowing a role to request a temporary endpoint on a random port > 1023 would allow this.

Hope to see TDE (preferred) or column encryption support on Azure SQL.

'Imote Desktop for Windows Azure' is RemoteDesktop for Windows Azrure with client Agent Service.
"Imote" is Japanese that becomes a meaning of "Younger sister" in English. It is called "Imote" to be friendly, lovely, to want to use in Japan, and to become it. This service arranges the agent who mounts UX as friendly as the younger sister on the client, and comes to be able to manage Windows Azure by NUI of the conversation and the gesture, etc.
Image Screenshot is:
http://blogs.msdn.com/cfs-file.ashx/__key/communityserver-blogs-components-weblogfiles/00-00-01-51-80-ap/6710.ap_5F00_img003.png

It would be great to have a CURRENT list of IP Ranges allocated to each Azure Datacenter.

The page should be on the Management Portal somewhere, e.g. under "Windows Azure Service Status: Running"
"Azure Current IP Ranges:...."

Being on an endorsed MS page would ease Client's concerns about punching holes to not recognised IP Addresses.

The ability to backup federation members or even better single atomic units is a must when using federation and is not supported at this time

It is good if you can give a provision to add start up tasks for persistent VMs. We can create on based VHD, while spinning up the VMs probably we might want to add some additional tools or softwares on the fly based on some business logics and conditions.

given a key (or perhaps delegate for encryption/decryption) at mount time, transparently encrypt/decrypt all read/writes against the mounted clouddrive so that stored blob data is encrypted transparently.

Microsoft should send an email alert to the Azure service administrators before applying Guest OS updates. If a developer do not mention Guest OS version in Azure config file then Microsoft keeps updating the VM with latest releases and patches which is fine, but Microsoft should send an email alert 1 or 2 days earlier before applying any update to the VM's.

Allow connection to VPN from multiple sites even from sites which are on Dynamic IP addressed to use azure as central site and others a branch offices

As well as allow the possibility of windows machines with direct access to connect to azure VPN

I wish MSFT could reserve a name to the last using live id ... at least for 72 hours ... to allow someone to feel comfortable about deleting a project name.

see: http://social.msdn.microsoft.com/Forums/en-US/windowsazure/thread/c42efcc4-925d-48d9-b71b-ad94e9a77dff/

Too often changes are being implemented onto the production Azure platform without the ability for devs to test / validate that the changes won't affect their production apps. Need to know what is in the pipeline, when it will be applied, and what to look for to evaluate the impact. Right now, we don't know something will affect us until our production site goes down.