Help in understanding Hinted Handoffs and Data replication in Cassandra

I initially just wanted to point at this as a comment, but my rep is too low. So excuse me if this is not a fully fledged answer. (Anyone feel free to edit or hint in comments)

Looking at the documentation at datastax it appears that data will be evently divided by nodes, depending on the "automagical" tokens. From my understanding it's simple as that a new vnode with new tokens just takes an even portion of each node, like all the other nodes already do. Hell, this is even what the doc says basically word by word:

Rebalancing a cluster is no longer necessary when adding or removing nodes. When a node joins the cluster, it assumes responsibility for an even portion of data from the other nodes in the cluster. If a node fails, the load is spread evenly across other nodes in the cluster.

And to address your question:

To further specify my question for virtual nodes does the cluster re-evaluate tokens for all nodes just because a new host has been added?

Sort of. The new node itself takes some partitions (evenly) from all other nodes. If you remove a node every other node will rebalance its partitions from the other nodes, making up for the lost node.

C stands for eventual consistent. Correct ?

Consistency in the CAP theorem is referring to strong consistency where every read receives the most recent write or an error. By default MongoDB drivers direct all reads & writes to the primary of a replica set, which is strongly consistent.

The CAP theorem asserts that a distributed system must choose between consistency and availability in the event of a network partition. MongoDB's replica set approach uses a single primary for write consistency (CP), while Cassandra's replication strategy favours write availability (AP). Strong consistency is not possible with a network partition because there could be a conflict if both sides of the partition update the same data. To maintain write availability AP database systems need a solution for conflict resolution, which is a separate consideration from eventual consistency.

However, CAP is a simplification of real-world behaviour: MongoDB and Cassandra both have tunable levels of consistency for reads and writes. For example: MongoDB has write concerns to determine the level of acknowledgement required for write operations, read preferences for routing requests to members of a replica set, and read concerns to control the recency, consistency, and isolation properties of data read from replica set and sharded deployments.

Eric Brewer, author of the CAP Theorem, revisited this in 2012 with a more nuanced take: CAP Twelve Years Later: How the "Rules" Have Changed.

3. Till master is re-elected, write requests need to wait and system is not available

There are no writes without a primary, but replica sets still have read availability. MongoDB 3.6 added a Retryable Writes feature which helps applications better handle replica set elections and transient network errors.

4. Once the previous node(node crashed in step 2) comes back, writes pending from that node are written back to slaves.

If the primary in a MongoDB replica set becomes unavailable, the remaining members of the replica set will elect a new primary if there is an eligible secondary and a quorum of voting members. In your example, the voting majority would be 2/3 members of your replica set. Any writes accepted by a former primary that were not written to a majority of replica set members will be rolled back (saved to disk) so the former primary resumes syncing from a state consistent with the history of the current primary.