Encrypt multiple messages with sodium_crypto_box_seal using the same public key

[doMynation]

Hi,

I'm a beginner in the realm of cryptography and I'd need some advice. I did quite a bit of research to familiarize myself with all the options available, and so far I narrowed it down to this fantastic library and sodium_crypto_box_seal() specifically. My use case is the following:

1. I need to generate a key pair (public/private), keep the public key and give the private key with a third party (and not keep any copy) such that I am only able to encrypt and not decrypt.
2. Over a period of a few days, I should be able to encrypt the data/records transmitted to me by other parties using the aforementioned public key and store it in my database as separate records.
3. Once the period is over, I should be able to take the private key (transmitted to me by the third party in bullet 1) and go over all the data/records that have been stored so far and decrypt them one by one.

Given the above, I have the following questions:

1. It's not clear to me what the underlying mechanism of sodium_crypto_box_seal() is, I'm assuming it's handling the nonce/IV/salt automatically for each encryption? Assuming that it the case, is it safe to use the same public key to encrypt many messages over time (bullet 2 above)?
2. Is the resulting ciphertext also containing information about what cipher/algorithm is used? According to this, the underlying implementation currently uses X25519 and XSalsa20-Poly1305, but what if those become obsolete in the future and better ciphers/algorithms become available? Will libsodium's sodium_crypto_box_seal() be updated to use them in a backward compatible fashion or would the library expose new functions in this case?
3. When the time comes to decrypt all the records accumulated (bullet 3), I might have tens of thousands of records to decrypt, one by one, using sodium_crypto_box_seal_open(). Is there some known construct/function I can use to speed up the process or should I just expect the operation to take a very long time?
4. Is there perhaps a better alternative for my use case?

Thank you very much for taking the time to read this.