Using driver with R libraries RJDBC and DBI crashes script

I have been using RJDBC and DBI ibraries in the past with the RPostgres::Postgres() driver without problems. But it requires authentication by userID and password. Recently all the data of interest has migrated to Andes 3.0 and is available only with IAM authentication (access_key_id, secret_access_key, session_token). It seems that your driver is the only one that accept this type of authentication. So, I am trying now to use it as in the script below. It works for a few minutes (I can display the head of the table for example), but the R session invariably ends up crashing with a `*** caught segfault *** address (nil), cause ‘memory not mapped’` error.
I run the script on a cloud desktop (c6a.4xlarge) running Amazon Linux 2 (x86_64 5.10) with R-4.4.2 with packages versions:
`RJDBC_0.2-10`, `rJava_1.0-11`, `DBI_1.2.3`.
Let me know if you need more details.

```
### MAIN ###

#Code to connect to Andes with JDBC#
#(See https://docs.aws.amazon.com/redshift/latest/mgmt/jdbc20-install.html for more info)

#General libraries
library(tidyverse)
library(magrittr)

#Specific libraries for connecting to DBs
#library(rJava)
library(RJDBC)
library(DBI)

fname_driver_jdbc <- "redshift-jdbc42-2.1.0.32.jar" #downloaded from https://docs.aws.amazon.com/redshift/latest/mgmt/jdbc20-download-driver.html
path_driver <- "/home/myself/R/drivers/"
fpath_driver_jdbc <- glue("{path_driver}{fname_driver_jdbc}")

driver_0 <- #From RJDBC library:
JDBC(driverClass = "com.amazon.redshift.Driver",
classPath = Sys.glob(fpath_driver_jdbc),
identifier.quote = "`")

region_0 <- "us-east-1"
db_name_0 <- "XXXX_data_prod"
host_0 <- "XXXX-data-redshift.abcdefghijklmnop-east-1.redshift.amazonaws.com"
port_0 <- 5438
clust_id_0 <- "XXXX-data-redshift"
url_0 <- glue("jdbc:redshift:iam://{host_0}")

profil_IDP <- "XXXX-data-blah"
##From `ada profile list`:
#Profile: XXXX-data-blah
#Provider: conduit
#Account: XXXXXXXXXXXX
#Role: IibsAdminAccess-DO-NOT-DELETE
#Partition: aws
dttm_exp <- fun_update_cred2(profil_IDP) #First run code from appendix

conn_andes <- #From DBI library
dbConnect(drv = driver_0,
url = url_0,
DBNAME = db_name_0,
PORT = port_0,
Region = region_0,
ClusterID = clust_id_0,
groupFederation = TRUE,
AccessKeyID = Sys.getenv("AWS_ACCESS_KEY_ID"),
SecretAccessKey = Sys.getenv("AWS_SECRET_ACCESS_KEY"),
SessionToken = Sys.getenv("AWS_SESSION_TOKEN"))

catalog_blah <- "XXXXX"
schema_blah <- "XXXXXX"
table_blah <- "xxxxxxxx"

db_blah <- tbl(conn_andes, in_catalog(catalog_blah, schema_blah, table_blah))

db_blah %>% head

#It works! But if I include that in a script, after a little while the script crashes.
#The error: "*** caught segfault *** address (nil), cause ‘memory not mapped’" hints toward a memory leak.

### APPENDIX ###

#Code for updating credentials#

#retrieves access_key, secret_key, and session_token from ada CLI
fun_aws_cred2 <- function(profil = profil_default) {
cmd_ada <- glue("ada credentials print --profile {profil}")
cred_json <- system(cmd_ada, intern = TRUE)
list_cred <- jsonlite::fromJSON(cred_json)
dttm_exp <- lubridate::as_datetime(list_cred$Expiration) #UTC (default)
return(list(aws_key_id = list_cred$AccessKeyId,
aws_key_secret = list_cred$SecretAccessKey,
aws_sess_token = list_cred$SessionToken,
aws_dttm_exp = dttm_exp))
}

#Saves access-key, secret-key, and session_token into environment variables and returns credential expiration datetime.
fun_update_cred2 <- function(profil = profil_default) {
aws_region <- fun_aws_region(profil) #This just retrieves region from ~/.aws/config (usually us-east-1)
aws_cred <- fun_aws_cred2(profil)
new_key <- aws_cred['aws_key_id']
dttm_exp <- aws_cred['aws_dttm_exp'] %>% as.numeric
Sys.setenv("AWS_DEFAULT_REGION" = aws_region,
"AWS_ACCESS_KEY_ID" = new_key,
"AWS_SECRET_ACCESS_KEY" = aws_cred['aws_key_secret'],
"AWS_SESSION_TOKEN" = aws_cred['aws_sess_token'])
return(lubridate::as_datetime(dttm_exp))
}
```

Start by reproducing the supplied RJDBC/DBI script with the Redshift JDBC JAR on the stated R and Amazon Linux environment. Inspect the interaction between RJDBC, rJava, DBI, IAM credentials, and the driver around the point where the session segfaults. Done means the trigger or failing boundary is identified and the script no longer crashes.

Written by the indexing model from the issue text.